Training in progress, epoch 4

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:34978cf045cd0d1b178d8c62ec0e0d5c2f6f5d8dc386e88dace48f8296af6954
+oid sha256:7ae31e24763dbc75728917d8a248eb9fc1d9c04902255a6b2137d0f66af4b99a
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:902f2d039c5b8a559159a7d8de0012fa2d09642cf087fff24a37701ec04415fd
+oid sha256:0a2d96b9ffbf89951d6ddcfec5af928bf425c5ef45ca719f06012e3adf96f6b4
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:983753954d232b8925e38c53ec7c83d4628a543c82030feb6f40a1a5fb94bc3f
+oid sha256:b1e3732b6eace8c58b70de54256c15eedab178eeb58a9a1a05bbeda998bc64fc
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:52d3fe29ac0b97ca4650270c83042a55ef57df3a80e138fed76b7036036f6fe9
+oid sha256:7e96478caae079ddffc77dde5ea3937c4d14e52c73ff18a8e825835bcd0cb043
 size 1089994880

trainer_log.jsonl

@@ -783,3 +783,194 @@
 {"current_steps": 783, "total_steps": 975, "loss": 0.8363, "lr": 9.093789995562083e-06, "epoch": 4.005115089514066, "percentage": 80.31, "elapsed_time": "11:24:10", "remaining_time": "2:47:46"}
 {"current_steps": 784, "total_steps": 975, "loss": 0.8234, "lr": 9.0030254658449e-06, "epoch": 4.010230179028133, "percentage": 80.41, "elapsed_time": "11:25:02", "remaining_time": "2:46:53"}
 {"current_steps": 785, "total_steps": 975, "loss": 0.8467, "lr": 8.91265869456503e-06, "epoch": 4.015345268542199, "percentage": 80.51, "elapsed_time": "11:25:52", "remaining_time": "2:46:00"}
+{"current_steps": 786, "total_steps": 975, "loss": 0.84, "lr": 8.822690841324237e-06, "epoch": 4.020460358056266, "percentage": 80.62, "elapsed_time": "11:26:42", "remaining_time": "2:45:07"}
+{"current_steps": 787, "total_steps": 975, "loss": 0.8301, "lr": 8.733123060605262e-06, "epoch": 4.025575447570333, "percentage": 80.72, "elapsed_time": "11:27:33", "remaining_time": "2:44:14"}
+{"current_steps": 788, "total_steps": 975, "loss": 0.8428, "lr": 8.643956501757072e-06, "epoch": 4.030690537084399, "percentage": 80.82, "elapsed_time": "11:28:24", "remaining_time": "2:43:22"}
+{"current_steps": 789, "total_steps": 975, "loss": 0.8323, "lr": 8.555192308980095e-06, "epoch": 4.035805626598465, "percentage": 80.92, "elapsed_time": "11:29:15", "remaining_time": "2:42:29"}
+{"current_steps": 790, "total_steps": 975, "loss": 0.8506, "lr": 8.466831621311496e-06, "epoch": 4.040920716112532, "percentage": 81.03, "elapsed_time": "11:30:07", "remaining_time": "2:41:36"}
+{"current_steps": 791, "total_steps": 975, "loss": 0.8325, "lr": 8.378875572610625e-06, "epoch": 4.046035805626598, "percentage": 81.13, "elapsed_time": "11:30:57", "remaining_time": "2:40:43"}
+{"current_steps": 792, "total_steps": 975, "loss": 0.8286, "lr": 8.29132529154442e-06, "epoch": 4.051150895140665, "percentage": 81.23, "elapsed_time": "11:31:48", "remaining_time": "2:39:51"}
+{"current_steps": 793, "total_steps": 975, "loss": 0.8512, "lr": 8.204181901572958e-06, "epoch": 4.056265984654732, "percentage": 81.33, "elapsed_time": "11:32:39", "remaining_time": "2:38:58"}
+{"current_steps": 794, "total_steps": 975, "loss": 0.814, "lr": 8.117446520934989e-06, "epoch": 4.061381074168798, "percentage": 81.44, "elapsed_time": "11:33:30", "remaining_time": "2:38:05"}
+{"current_steps": 795, "total_steps": 975, "loss": 0.843, "lr": 8.031120262633641e-06, "epoch": 4.0664961636828645, "percentage": 81.54, "elapsed_time": "11:34:23", "remaining_time": "2:37:13"}
+{"current_steps": 796, "total_steps": 975, "loss": 0.8584, "lr": 7.945204234422124e-06, "epoch": 4.071611253196931, "percentage": 81.64, "elapsed_time": "11:35:21", "remaining_time": "2:36:22"}
+{"current_steps": 797, "total_steps": 975, "loss": 0.8642, "lr": 7.859699538789466e-06, "epoch": 4.076726342710997, "percentage": 81.74, "elapsed_time": "11:36:19", "remaining_time": "2:35:30"}
+{"current_steps": 798, "total_steps": 975, "loss": 0.8422, "lr": 7.774607272946465e-06, "epoch": 4.081841432225064, "percentage": 81.85, "elapsed_time": "11:37:09", "remaining_time": "2:34:38"}
+{"current_steps": 799, "total_steps": 975, "loss": 0.8281, "lr": 7.68992852881147e-06, "epoch": 4.086956521739131, "percentage": 81.95, "elapsed_time": "11:37:59", "remaining_time": "2:33:45"}
+{"current_steps": 800, "total_steps": 975, "loss": 0.8288, "lr": 7.605664392996544e-06, "epoch": 4.092071611253197, "percentage": 82.05, "elapsed_time": "11:38:51", "remaining_time": "2:32:52"}
+{"current_steps": 801, "total_steps": 975, "loss": 0.835, "lr": 7.521815946793354e-06, "epoch": 4.0971867007672635, "percentage": 82.15, "elapsed_time": "11:39:42", "remaining_time": "2:31:59"}
+{"current_steps": 802, "total_steps": 975, "loss": 0.8387, "lr": 7.438384266159384e-06, "epoch": 4.10230179028133, "percentage": 82.26, "elapsed_time": "11:40:34", "remaining_time": "2:31:07"}
+{"current_steps": 803, "total_steps": 975, "loss": 0.8421, "lr": 7.355370421704124e-06, "epoch": 4.107416879795396, "percentage": 82.36, "elapsed_time": "11:41:25", "remaining_time": "2:30:14"}
+{"current_steps": 804, "total_steps": 975, "loss": 0.8275, "lr": 7.272775478675318e-06, "epoch": 4.112531969309463, "percentage": 82.46, "elapsed_time": "11:42:16", "remaining_time": "2:29:21"}
+{"current_steps": 805, "total_steps": 975, "loss": 0.8428, "lr": 7.190600496945297e-06, "epoch": 4.117647058823529, "percentage": 82.56, "elapsed_time": "11:43:07", "remaining_time": "2:28:29"}
+{"current_steps": 806, "total_steps": 975, "loss": 0.8278, "lr": 7.108846530997349e-06, "epoch": 4.122762148337596, "percentage": 82.67, "elapsed_time": "11:43:58", "remaining_time": "2:27:36"}
+{"current_steps": 807, "total_steps": 975, "loss": 0.8457, "lr": 7.027514629912256e-06, "epoch": 4.127877237851663, "percentage": 82.77, "elapsed_time": "11:44:49", "remaining_time": "2:26:43"}
+{"current_steps": 808, "total_steps": 975, "loss": 0.8439, "lr": 6.946605837354781e-06, "epoch": 4.132992327365729, "percentage": 82.87, "elapsed_time": "11:45:40", "remaining_time": "2:25:51"}
+{"current_steps": 809, "total_steps": 975, "loss": 0.8347, "lr": 6.866121191560262e-06, "epoch": 4.138107416879795, "percentage": 82.97, "elapsed_time": "11:46:31", "remaining_time": "2:24:58"}
+{"current_steps": 810, "total_steps": 975, "loss": 0.8345, "lr": 6.786061725321347e-06, "epoch": 4.143222506393862, "percentage": 83.08, "elapsed_time": "11:47:24", "remaining_time": "2:24:06"}
+{"current_steps": 811, "total_steps": 975, "loss": 0.8424, "lr": 6.7064284659746724e-06, "epoch": 4.148337595907928, "percentage": 83.18, "elapsed_time": "11:48:19", "remaining_time": "2:23:14"}
+{"current_steps": 812, "total_steps": 975, "loss": 0.8424, "lr": 6.627222435387755e-06, "epoch": 4.153452685421995, "percentage": 83.28, "elapsed_time": "11:49:17", "remaining_time": "2:22:22"}
+{"current_steps": 813, "total_steps": 975, "loss": 0.8396, "lr": 6.548444649945818e-06, "epoch": 4.158567774936062, "percentage": 83.38, "elapsed_time": "11:50:11", "remaining_time": "2:21:30"}
+{"current_steps": 814, "total_steps": 975, "loss": 0.8469, "lr": 6.4700961205387625e-06, "epoch": 4.163682864450128, "percentage": 83.49, "elapsed_time": "11:51:02", "remaining_time": "2:20:38"}
+{"current_steps": 815, "total_steps": 975, "loss": 0.8505, "lr": 6.392177852548216e-06, "epoch": 4.168797953964194, "percentage": 83.59, "elapsed_time": "11:51:53", "remaining_time": "2:19:45"}
+{"current_steps": 816, "total_steps": 975, "loss": 0.8502, "lr": 6.3146908458346345e-06, "epoch": 4.173913043478261, "percentage": 83.69, "elapsed_time": "11:52:44", "remaining_time": "2:18:52"}
+{"current_steps": 817, "total_steps": 975, "loss": 0.8353, "lr": 6.237636094724409e-06, "epoch": 4.179028132992327, "percentage": 83.79, "elapsed_time": "11:53:35", "remaining_time": "2:18:00"}
+{"current_steps": 818, "total_steps": 975, "loss": 0.8589, "lr": 6.161014587997196e-06, "epoch": 4.1841432225063935, "percentage": 83.9, "elapsed_time": "11:54:26", "remaining_time": "2:17:07"}
+{"current_steps": 819, "total_steps": 975, "loss": 0.8415, "lr": 6.084827308873169e-06, "epoch": 4.189258312020461, "percentage": 84.0, "elapsed_time": "11:55:17", "remaining_time": "2:16:14"}
+{"current_steps": 820, "total_steps": 975, "loss": 0.834, "lr": 6.009075235000437e-06, "epoch": 4.194373401534527, "percentage": 84.1, "elapsed_time": "11:56:08", "remaining_time": "2:15:22"}
+{"current_steps": 821, "total_steps": 975, "loss": 0.8334, "lr": 5.9337593384424506e-06, "epoch": 4.1994884910485935, "percentage": 84.21, "elapsed_time": "11:56:59", "remaining_time": "2:14:29"}
+{"current_steps": 822, "total_steps": 975, "loss": 0.8334, "lr": 5.858880585665563e-06, "epoch": 4.20460358056266, "percentage": 84.31, "elapsed_time": "11:57:49", "remaining_time": "2:13:36"}
+{"current_steps": 823, "total_steps": 975, "loss": 0.8331, "lr": 5.784439937526674e-06, "epoch": 4.209718670076726, "percentage": 84.41, "elapsed_time": "11:58:40", "remaining_time": "2:12:43"}
+{"current_steps": 824, "total_steps": 975, "loss": 0.8298, "lr": 5.710438349260781e-06, "epoch": 4.2148337595907925, "percentage": 84.51, "elapsed_time": "11:59:31", "remaining_time": "2:11:51"}
+{"current_steps": 825, "total_steps": 975, "loss": 0.8216, "lr": 5.6368767704688425e-06, "epoch": 4.21994884910486, "percentage": 84.62, "elapsed_time": "12:00:22", "remaining_time": "2:10:58"}
+{"current_steps": 826, "total_steps": 975, "loss": 0.8406, "lr": 5.563756145105501e-06, "epoch": 4.225063938618926, "percentage": 84.72, "elapsed_time": "12:01:16", "remaining_time": "2:10:06"}
+{"current_steps": 827, "total_steps": 975, "loss": 0.8425, "lr": 5.491077411467038e-06, "epoch": 4.2301790281329925, "percentage": 84.82, "elapsed_time": "12:02:14", "remaining_time": "2:09:15"}
+{"current_steps": 828, "total_steps": 975, "loss": 0.8517, "lr": 5.418841502179315e-06, "epoch": 4.235294117647059, "percentage": 84.92, "elapsed_time": "12:03:12", "remaining_time": "2:08:23"}
+{"current_steps": 829, "total_steps": 975, "loss": 0.8434, "lr": 5.347049344185746e-06, "epoch": 4.240409207161125, "percentage": 85.03, "elapsed_time": "12:04:04", "remaining_time": "2:07:31"}
+{"current_steps": 830, "total_steps": 975, "loss": 0.8403, "lr": 5.2757018587355066e-06, "epoch": 4.245524296675192, "percentage": 85.13, "elapsed_time": "12:04:54", "remaining_time": "2:06:38"}
+{"current_steps": 831, "total_steps": 975, "loss": 0.8477, "lr": 5.20479996137163e-06, "epoch": 4.250639386189258, "percentage": 85.23, "elapsed_time": "12:05:44", "remaining_time": "2:05:45"}
+{"current_steps": 832, "total_steps": 975, "loss": 0.8307, "lr": 5.134344561919306e-06, "epoch": 4.255754475703325, "percentage": 85.33, "elapsed_time": "12:06:35", "remaining_time": "2:04:53"}
+{"current_steps": 833, "total_steps": 975, "loss": 0.835, "lr": 5.064336564474168e-06, "epoch": 4.260869565217392, "percentage": 85.44, "elapsed_time": "12:07:27", "remaining_time": "2:04:00"}
+{"current_steps": 834, "total_steps": 975, "loss": 0.8229, "lr": 4.9947768673907025e-06, "epoch": 4.265984654731458, "percentage": 85.54, "elapsed_time": "12:08:17", "remaining_time": "2:03:07"}
+{"current_steps": 835, "total_steps": 975, "loss": 0.8435, "lr": 4.925666363270782e-06, "epoch": 4.271099744245524, "percentage": 85.64, "elapsed_time": "12:09:09", "remaining_time": "2:02:15"}
+{"current_steps": 836, "total_steps": 975, "loss": 0.8294, "lr": 4.857005938952095e-06, "epoch": 4.276214833759591, "percentage": 85.74, "elapsed_time": "12:10:00", "remaining_time": "2:01:22"}
+{"current_steps": 837, "total_steps": 975, "loss": 0.8153, "lr": 4.788796475496873e-06, "epoch": 4.281329923273657, "percentage": 85.85, "elapsed_time": "12:10:52", "remaining_time": "2:00:30"}
+{"current_steps": 838, "total_steps": 975, "loss": 0.8386, "lr": 4.721038848180515e-06, "epoch": 4.286445012787723, "percentage": 85.95, "elapsed_time": "12:11:43", "remaining_time": "1:59:37"}
+{"current_steps": 839, "total_steps": 975, "loss": 0.8468, "lr": 4.6537339264803994e-06, "epoch": 4.291560102301791, "percentage": 86.05, "elapsed_time": "12:12:34", "remaining_time": "1:58:44"}
+{"current_steps": 840, "total_steps": 975, "loss": 0.8423, "lr": 4.586882574064708e-06, "epoch": 4.296675191815857, "percentage": 86.15, "elapsed_time": "12:13:25", "remaining_time": "1:57:52"}
+{"current_steps": 841, "total_steps": 975, "loss": 0.8461, "lr": 4.5204856487813275e-06, "epoch": 4.301790281329923, "percentage": 86.26, "elapsed_time": "12:14:17", "remaining_time": "1:56:59"}
+{"current_steps": 842, "total_steps": 975, "loss": 0.8587, "lr": 4.4545440026468744e-06, "epoch": 4.30690537084399, "percentage": 86.36, "elapsed_time": "12:15:12", "remaining_time": "1:56:07"}
+{"current_steps": 843, "total_steps": 975, "loss": 0.8374, "lr": 4.389058481835751e-06, "epoch": 4.312020460358056, "percentage": 86.46, "elapsed_time": "12:16:11", "remaining_time": "1:55:16"}
+{"current_steps": 844, "total_steps": 975, "loss": 0.82, "lr": 4.324029926669262e-06, "epoch": 4.3171355498721224, "percentage": 86.56, "elapsed_time": "12:17:06", "remaining_time": "1:54:24"}
+{"current_steps": 845, "total_steps": 975, "loss": 0.85, "lr": 4.259459171604863e-06, "epoch": 4.322250639386189, "percentage": 86.67, "elapsed_time": "12:17:57", "remaining_time": "1:53:31"}
+{"current_steps": 846, "total_steps": 975, "loss": 0.8327, "lr": 4.195347045225444e-06, "epoch": 4.327365728900256, "percentage": 86.77, "elapsed_time": "12:18:48", "remaining_time": "1:52:39"}
+{"current_steps": 847, "total_steps": 975, "loss": 0.8504, "lr": 4.131694370228703e-06, "epoch": 4.332480818414322, "percentage": 86.87, "elapsed_time": "12:19:39", "remaining_time": "1:51:46"}
+{"current_steps": 848, "total_steps": 975, "loss": 0.843, "lr": 4.068501963416549e-06, "epoch": 4.337595907928389, "percentage": 86.97, "elapsed_time": "12:20:30", "remaining_time": "1:50:54"}
+{"current_steps": 849, "total_steps": 975, "loss": 0.8213, "lr": 4.005770635684671e-06, "epoch": 4.342710997442455, "percentage": 87.08, "elapsed_time": "12:21:21", "remaining_time": "1:50:01"}
+{"current_steps": 850, "total_steps": 975, "loss": 0.8435, "lr": 3.943501192012118e-06, "epoch": 4.3478260869565215, "percentage": 87.18, "elapsed_time": "12:22:11", "remaining_time": "1:49:08"}
+{"current_steps": 851, "total_steps": 975, "loss": 0.8285, "lr": 3.881694431450953e-06, "epoch": 4.352941176470588, "percentage": 87.28, "elapsed_time": "12:23:02", "remaining_time": "1:48:16"}
+{"current_steps": 852, "total_steps": 975, "loss": 0.8463, "lr": 3.8203511471160255e-06, "epoch": 4.358056265984655, "percentage": 87.38, "elapsed_time": "12:23:53", "remaining_time": "1:47:23"}
+{"current_steps": 853, "total_steps": 975, "loss": 0.8482, "lr": 3.759472126174761e-06, "epoch": 4.3631713554987215, "percentage": 87.49, "elapsed_time": "12:24:43", "remaining_time": "1:46:30"}
+{"current_steps": 854, "total_steps": 975, "loss": 0.8354, "lr": 3.699058149837087e-06, "epoch": 4.368286445012788, "percentage": 87.59, "elapsed_time": "12:25:35", "remaining_time": "1:45:38"}
+{"current_steps": 855, "total_steps": 975, "loss": 0.8502, "lr": 3.639109993345411e-06, "epoch": 4.373401534526854, "percentage": 87.69, "elapsed_time": "12:26:26", "remaining_time": "1:44:45"}
+{"current_steps": 856, "total_steps": 975, "loss": 0.8608, "lr": 3.5796284259646342e-06, "epoch": 4.378516624040921, "percentage": 87.79, "elapsed_time": "12:27:17", "remaining_time": "1:43:53"}
+{"current_steps": 857, "total_steps": 975, "loss": 0.8416, "lr": 3.5206142109723306e-06, "epoch": 4.383631713554987, "percentage": 87.9, "elapsed_time": "12:28:10", "remaining_time": "1:43:01"}
+{"current_steps": 858, "total_steps": 975, "loss": 0.8433, "lr": 3.4620681056489213e-06, "epoch": 4.388746803069053, "percentage": 88.0, "elapsed_time": "12:29:08", "remaining_time": "1:42:09"}
+{"current_steps": 859, "total_steps": 975, "loss": 0.8439, "lr": 3.403990861267974e-06, "epoch": 4.3938618925831205, "percentage": 88.1, "elapsed_time": "12:30:05", "remaining_time": "1:41:17"}
+{"current_steps": 860, "total_steps": 975, "loss": 0.8376, "lr": 3.3463832230865355e-06, "epoch": 4.398976982097187, "percentage": 88.21, "elapsed_time": "12:30:57", "remaining_time": "1:40:25"}
+{"current_steps": 861, "total_steps": 975, "loss": 0.8604, "lr": 3.2892459303355806e-06, "epoch": 4.404092071611253, "percentage": 88.31, "elapsed_time": "12:31:48", "remaining_time": "1:39:32"}
+{"current_steps": 862, "total_steps": 975, "loss": 0.8439, "lr": 3.232579716210573e-06, "epoch": 4.40920716112532, "percentage": 88.41, "elapsed_time": "12:32:38", "remaining_time": "1:38:39"}
+{"current_steps": 863, "total_steps": 975, "loss": 0.8383, "lr": 3.176385307861969e-06, "epoch": 4.414322250639386, "percentage": 88.51, "elapsed_time": "12:33:30", "remaining_time": "1:37:47"}
+{"current_steps": 864, "total_steps": 975, "loss": 0.8304, "lr": 3.120663426385968e-06, "epoch": 4.419437340153452, "percentage": 88.62, "elapsed_time": "12:34:21", "remaining_time": "1:36:54"}
+{"current_steps": 865, "total_steps": 975, "loss": 0.8507, "lr": 3.065414786815191e-06, "epoch": 4.42455242966752, "percentage": 88.72, "elapsed_time": "12:35:13", "remaining_time": "1:36:02"}
+{"current_steps": 866, "total_steps": 975, "loss": 0.8224, "lr": 3.0106400981095675e-06, "epoch": 4.429667519181586, "percentage": 88.82, "elapsed_time": "12:36:04", "remaining_time": "1:35:09"}
+{"current_steps": 867, "total_steps": 975, "loss": 0.8348, "lr": 2.9563400631472005e-06, "epoch": 4.434782608695652, "percentage": 88.92, "elapsed_time": "12:36:55", "remaining_time": "1:34:17"}
+{"current_steps": 868, "total_steps": 975, "loss": 0.8591, "lr": 2.902515378715345e-06, "epoch": 4.439897698209719, "percentage": 89.03, "elapsed_time": "12:37:46", "remaining_time": "1:33:24"}
+{"current_steps": 869, "total_steps": 975, "loss": 0.8355, "lr": 2.8491667355014896e-06, "epoch": 4.445012787723785, "percentage": 89.13, "elapsed_time": "12:38:37", "remaining_time": "1:32:32"}
+{"current_steps": 870, "total_steps": 975, "loss": 0.8313, "lr": 2.7962948180844817e-06, "epoch": 4.450127877237851, "percentage": 89.23, "elapsed_time": "12:39:28", "remaining_time": "1:31:39"}
+{"current_steps": 871, "total_steps": 975, "loss": 0.8466, "lr": 2.7439003049257286e-06, "epoch": 4.455242966751918, "percentage": 89.33, "elapsed_time": "12:40:19", "remaining_time": "1:30:47"}
+{"current_steps": 872, "total_steps": 975, "loss": 0.8319, "lr": 2.6919838683605236e-06, "epoch": 4.460358056265985, "percentage": 89.44, "elapsed_time": "12:41:11", "remaining_time": "1:29:54"}
+{"current_steps": 873, "total_steps": 975, "loss": 0.8452, "lr": 2.6405461745893756e-06, "epoch": 4.465473145780051, "percentage": 89.54, "elapsed_time": "12:42:07", "remaining_time": "1:29:02"}
+{"current_steps": 874, "total_steps": 975, "loss": 0.8365, "lr": 2.5895878836695198e-06, "epoch": 4.470588235294118, "percentage": 89.64, "elapsed_time": "12:43:05", "remaining_time": "1:28:10"}
+{"current_steps": 875, "total_steps": 975, "loss": 0.8556, "lr": 2.539109649506384e-06, "epoch": 4.475703324808184, "percentage": 89.74, "elapsed_time": "12:44:00", "remaining_time": "1:27:18"}
+{"current_steps": 876, "total_steps": 975, "loss": 0.8382, "lr": 2.4891121198452297e-06, "epoch": 4.4808184143222505, "percentage": 89.85, "elapsed_time": "12:44:51", "remaining_time": "1:26:26"}
+{"current_steps": 877, "total_steps": 975, "loss": 0.8539, "lr": 2.439595936262844e-06, "epoch": 4.485933503836317, "percentage": 89.95, "elapsed_time": "12:45:43", "remaining_time": "1:25:33"}
+{"current_steps": 878, "total_steps": 975, "loss": 0.8403, "lr": 2.3905617341593023e-06, "epoch": 4.491048593350383, "percentage": 90.05, "elapsed_time": "12:46:33", "remaining_time": "1:24:41"}
+{"current_steps": 879, "total_steps": 975, "loss": 0.8332, "lr": 2.3420101427498134e-06, "epoch": 4.4961636828644505, "percentage": 90.15, "elapsed_time": "12:47:25", "remaining_time": "1:23:48"}
+{"current_steps": 880, "total_steps": 975, "loss": 0.837, "lr": 2.293941785056619e-06, "epoch": 4.501278772378517, "percentage": 90.26, "elapsed_time": "12:48:16", "remaining_time": "1:22:56"}
+{"current_steps": 881, "total_steps": 975, "loss": 0.8566, "lr": 2.246357277901052e-06, "epoch": 4.506393861892583, "percentage": 90.36, "elapsed_time": "12:49:07", "remaining_time": "1:22:03"}
+{"current_steps": 882, "total_steps": 975, "loss": 0.8385, "lr": 2.1992572318955885e-06, "epoch": 4.5115089514066495, "percentage": 90.46, "elapsed_time": "12:49:57", "remaining_time": "1:21:11"}
+{"current_steps": 883, "total_steps": 975, "loss": 0.836, "lr": 2.1526422514360013e-06, "epoch": 4.516624040920716, "percentage": 90.56, "elapsed_time": "12:50:49", "remaining_time": "1:20:18"}
+{"current_steps": 884, "total_steps": 975, "loss": 0.8583, "lr": 2.106512934693634e-06, "epoch": 4.521739130434782, "percentage": 90.67, "elapsed_time": "12:51:39", "remaining_time": "1:19:26"}
+{"current_steps": 885, "total_steps": 975, "loss": 0.8312, "lr": 2.060869873607714e-06, "epoch": 4.526854219948849, "percentage": 90.77, "elapsed_time": "12:52:30", "remaining_time": "1:18:33"}
+{"current_steps": 886, "total_steps": 975, "loss": 0.8334, "lr": 2.015713653877733e-06, "epoch": 4.531969309462916, "percentage": 90.87, "elapsed_time": "12:53:21", "remaining_time": "1:17:41"}
+{"current_steps": 887, "total_steps": 975, "loss": 0.8288, "lr": 1.971044854955979e-06, "epoch": 4.537084398976982, "percentage": 90.97, "elapsed_time": "12:54:13", "remaining_time": "1:16:48"}
+{"current_steps": 888, "total_steps": 975, "loss": 0.826, "lr": 1.9268640500400426e-06, "epoch": 4.542199488491049, "percentage": 91.08, "elapsed_time": "12:55:06", "remaining_time": "1:15:56"}
+{"current_steps": 889, "total_steps": 975, "loss": 0.8492, "lr": 1.8831718060655335e-06, "epoch": 4.547314578005115, "percentage": 91.18, "elapsed_time": "12:56:03", "remaining_time": "1:15:04"}
+{"current_steps": 890, "total_steps": 975, "loss": 0.834, "lr": 1.839968683698725e-06, "epoch": 4.552429667519181, "percentage": 91.28, "elapsed_time": "12:57:01", "remaining_time": "1:14:12"}
+{"current_steps": 891, "total_steps": 975, "loss": 0.8486, "lr": 1.7972552373294184e-06, "epoch": 4.557544757033249, "percentage": 91.38, "elapsed_time": "12:57:53", "remaining_time": "1:13:20"}
+{"current_steps": 892, "total_steps": 975, "loss": 0.8517, "lr": 1.7550320150638001e-06, "epoch": 4.562659846547315, "percentage": 91.49, "elapsed_time": "12:58:44", "remaining_time": "1:12:27"}
+{"current_steps": 893, "total_steps": 975, "loss": 0.821, "lr": 1.7132995587174317e-06, "epoch": 4.567774936061381, "percentage": 91.59, "elapsed_time": "12:59:35", "remaining_time": "1:11:35"}
+{"current_steps": 894, "total_steps": 975, "loss": 0.8289, "lr": 1.672058403808281e-06, "epoch": 4.572890025575448, "percentage": 91.69, "elapsed_time": "13:00:26", "remaining_time": "1:10:42"}
+{"current_steps": 895, "total_steps": 975, "loss": 0.855, "lr": 1.6313090795498253e-06, "epoch": 4.578005115089514, "percentage": 91.79, "elapsed_time": "13:01:17", "remaining_time": "1:09:50"}
+{"current_steps": 896, "total_steps": 975, "loss": 0.839, "lr": 1.5910521088443242e-06, "epoch": 4.58312020460358, "percentage": 91.9, "elapsed_time": "13:02:08", "remaining_time": "1:08:57"}
+{"current_steps": 897, "total_steps": 975, "loss": 0.8387, "lr": 1.5512880082760595e-06, "epoch": 4.588235294117647, "percentage": 92.0, "elapsed_time": "13:02:59", "remaining_time": "1:08:05"}
+{"current_steps": 898, "total_steps": 975, "loss": 0.8296, "lr": 1.5120172881047057e-06, "epoch": 4.593350383631714, "percentage": 92.1, "elapsed_time": "13:03:51", "remaining_time": "1:07:12"}
+{"current_steps": 899, "total_steps": 975, "loss": 0.8326, "lr": 1.4732404522588195e-06, "epoch": 4.59846547314578, "percentage": 92.21, "elapsed_time": "13:04:41", "remaining_time": "1:06:20"}
+{"current_steps": 900, "total_steps": 975, "loss": 0.8431, "lr": 1.4349579983293204e-06, "epoch": 4.603580562659847, "percentage": 92.31, "elapsed_time": "13:05:32", "remaining_time": "1:05:27"}
+{"current_steps": 901, "total_steps": 975, "loss": 0.8167, "lr": 1.3971704175631807e-06, "epoch": 4.608695652173913, "percentage": 92.41, "elapsed_time": "13:06:23", "remaining_time": "1:04:35"}
+{"current_steps": 902, "total_steps": 975, "loss": 0.8252, "lr": 1.3598781948570339e-06, "epoch": 4.6138107416879794, "percentage": 92.51, "elapsed_time": "13:07:14", "remaining_time": "1:03:42"}
+{"current_steps": 903, "total_steps": 975, "loss": 0.8286, "lr": 1.3230818087510077e-06, "epoch": 4.618925831202046, "percentage": 92.62, "elapsed_time": "13:08:05", "remaining_time": "1:02:50"}
+{"current_steps": 904, "total_steps": 975, "loss": 0.8576, "lr": 1.2867817314225771e-06, "epoch": 4.624040920716112, "percentage": 92.72, "elapsed_time": "13:09:01", "remaining_time": "1:01:58"}
+{"current_steps": 905, "total_steps": 975, "loss": 0.8454, "lr": 1.2509784286804938e-06, "epoch": 4.629156010230179, "percentage": 92.82, "elapsed_time": "13:10:00", "remaining_time": "1:01:06"}
+{"current_steps": 906, "total_steps": 975, "loss": 0.8471, "lr": 1.2156723599588172e-06, "epoch": 4.634271099744246, "percentage": 92.92, "elapsed_time": "13:10:53", "remaining_time": "1:00:14"}
+{"current_steps": 907, "total_steps": 975, "loss": 0.8383, "lr": 1.1808639783110042e-06, "epoch": 4.639386189258312, "percentage": 93.03, "elapsed_time": "13:11:44", "remaining_time": "0:59:21"}
+{"current_steps": 908, "total_steps": 975, "loss": 0.8621, "lr": 1.146553730404123e-06, "epoch": 4.6445012787723785, "percentage": 93.13, "elapsed_time": "13:12:36", "remaining_time": "0:58:29"}
+{"current_steps": 909, "total_steps": 975, "loss": 0.8527, "lr": 1.1127420565131008e-06, "epoch": 4.649616368286445, "percentage": 93.23, "elapsed_time": "13:13:27", "remaining_time": "0:57:36"}
+{"current_steps": 910, "total_steps": 975, "loss": 0.8243, "lr": 1.0794293905150632e-06, "epoch": 4.654731457800511, "percentage": 93.33, "elapsed_time": "13:14:18", "remaining_time": "0:56:44"}
+{"current_steps": 911, "total_steps": 975, "loss": 0.8446, "lr": 1.0466161598837998e-06, "epoch": 4.659846547314578, "percentage": 93.44, "elapsed_time": "13:15:08", "remaining_time": "0:55:51"}
+{"current_steps": 912, "total_steps": 975, "loss": 0.8224, "lr": 1.0143027856842536e-06, "epoch": 4.664961636828645, "percentage": 93.54, "elapsed_time": "13:15:59", "remaining_time": "0:54:59"}
+{"current_steps": 913, "total_steps": 975, "loss": 0.8334, "lr": 9.82489682567125e-07, "epoch": 4.670076726342711, "percentage": 93.64, "elapsed_time": "13:16:50", "remaining_time": "0:54:06"}
+{"current_steps": 914, "total_steps": 975, "loss": 0.8275, "lr": 9.511772587635559e-07, "epoch": 4.675191815856778, "percentage": 93.74, "elapsed_time": "13:17:41", "remaining_time": "0:53:14"}
+{"current_steps": 915, "total_steps": 975, "loss": 0.8423, "lr": 9.203659160798728e-07, "epoch": 4.680306905370844, "percentage": 93.85, "elapsed_time": "13:18:32", "remaining_time": "0:52:21"}
+{"current_steps": 916, "total_steps": 975, "loss": 0.8486, "lr": 8.900560498924605e-07, "epoch": 4.68542199488491, "percentage": 93.95, "elapsed_time": "13:19:23", "remaining_time": "0:51:29"}
+{"current_steps": 917, "total_steps": 975, "loss": 0.8559, "lr": 8.60248049142669e-07, "epoch": 4.690537084398977, "percentage": 94.05, "elapsed_time": "13:20:14", "remaining_time": "0:50:36"}
+{"current_steps": 918, "total_steps": 975, "loss": 0.8554, "lr": 8.309422963318137e-07, "epoch": 4.695652173913043, "percentage": 94.15, "elapsed_time": "13:21:05", "remaining_time": "0:49:44"}
+{"current_steps": 919, "total_steps": 975, "loss": 0.8303, "lr": 8.021391675162892e-07, "epoch": 4.70076726342711, "percentage": 94.26, "elapsed_time": "13:21:58", "remaining_time": "0:48:52"}
+{"current_steps": 920, "total_steps": 975, "loss": 0.8401, "lr": 7.738390323027389e-07, "epoch": 4.705882352941177, "percentage": 94.36, "elapsed_time": "13:22:57", "remaining_time": "0:48:00"}
+{"current_steps": 921, "total_steps": 975, "loss": 0.8271, "lr": 7.460422538432976e-07, "epoch": 4.710997442455243, "percentage": 94.46, "elapsed_time": "13:23:55", "remaining_time": "0:47:08"}
+{"current_steps": 922, "total_steps": 975, "loss": 0.8381, "lr": 7.187491888309384e-07, "epoch": 4.716112531969309, "percentage": 94.56, "elapsed_time": "13:24:47", "remaining_time": "0:46:15"}
+{"current_steps": 923, "total_steps": 975, "loss": 0.8369, "lr": 6.919601874949066e-07, "epoch": 4.721227621483376, "percentage": 94.67, "elapsed_time": "13:25:39", "remaining_time": "0:45:23"}
+{"current_steps": 924, "total_steps": 975, "loss": 0.8524, "lr": 6.656755935962178e-07, "epoch": 4.726342710997442, "percentage": 94.77, "elapsed_time": "13:26:30", "remaining_time": "0:44:30"}
+{"current_steps": 925, "total_steps": 975, "loss": 0.8499, "lr": 6.398957444232335e-07, "epoch": 4.731457800511509, "percentage": 94.87, "elapsed_time": "13:27:21", "remaining_time": "0:43:38"}
+{"current_steps": 926, "total_steps": 975, "loss": 0.8338, "lr": 6.146209707873585e-07, "epoch": 4.736572890025576, "percentage": 94.97, "elapsed_time": "13:28:12", "remaining_time": "0:42:46"}
+{"current_steps": 927, "total_steps": 975, "loss": 0.8643, "lr": 5.89851597018769e-07, "epoch": 4.741687979539642, "percentage": 95.08, "elapsed_time": "13:29:03", "remaining_time": "0:41:53"}
+{"current_steps": 928, "total_steps": 975, "loss": 0.8239, "lr": 5.655879409622866e-07, "epoch": 4.746803069053708, "percentage": 95.18, "elapsed_time": "13:29:52", "remaining_time": "0:41:01"}
+{"current_steps": 929, "total_steps": 975, "loss": 0.8455, "lr": 5.418303139732573e-07, "epoch": 4.751918158567775, "percentage": 95.28, "elapsed_time": "13:30:43", "remaining_time": "0:40:08"}
+{"current_steps": 930, "total_steps": 975, "loss": 0.8446, "lr": 5.185790209135722e-07, "epoch": 4.757033248081841, "percentage": 95.38, "elapsed_time": "13:31:35", "remaining_time": "0:39:16"}
+{"current_steps": 931, "total_steps": 975, "loss": 0.8441, "lr": 4.958343601477822e-07, "epoch": 4.762148337595908, "percentage": 95.49, "elapsed_time": "13:32:26", "remaining_time": "0:38:23"}
+{"current_steps": 932, "total_steps": 975, "loss": 0.8549, "lr": 4.7359662353922974e-07, "epoch": 4.767263427109975, "percentage": 95.59, "elapsed_time": "13:33:17", "remaining_time": "0:37:31"}
+{"current_steps": 933, "total_steps": 975, "loss": 0.8421, "lr": 4.518660964463273e-07, "epoch": 4.772378516624041, "percentage": 95.69, "elapsed_time": "13:34:07", "remaining_time": "0:36:38"}
+{"current_steps": 934, "total_steps": 975, "loss": 0.8308, "lr": 4.306430577188758e-07, "epoch": 4.7774936061381075, "percentage": 95.79, "elapsed_time": "13:34:58", "remaining_time": "0:35:46"}
+{"current_steps": 935, "total_steps": 975, "loss": 0.8401, "lr": 4.0992777969452113e-07, "epoch": 4.782608695652174, "percentage": 95.9, "elapsed_time": "13:35:54", "remaining_time": "0:34:54"}
+{"current_steps": 936, "total_steps": 975, "loss": 0.8526, "lr": 3.8972052819522324e-07, "epoch": 4.78772378516624, "percentage": 96.0, "elapsed_time": "13:36:53", "remaining_time": "0:34:02"}
+{"current_steps": 937, "total_steps": 975, "loss": 0.8216, "lr": 3.7002156252386344e-07, "epoch": 4.792838874680307, "percentage": 96.1, "elapsed_time": "13:37:48", "remaining_time": "0:33:09"}
+{"current_steps": 938, "total_steps": 975, "loss": 0.8311, "lr": 3.508311354609051e-07, "epoch": 4.797953964194374, "percentage": 96.21, "elapsed_time": "13:38:39", "remaining_time": "0:32:17"}
+{"current_steps": 939, "total_steps": 975, "loss": 0.8444, "lr": 3.3214949326117354e-07, "epoch": 4.80306905370844, "percentage": 96.31, "elapsed_time": "13:39:29", "remaining_time": "0:31:25"}
+{"current_steps": 940, "total_steps": 975, "loss": 0.8246, "lr": 3.139768756506634e-07, "epoch": 4.8081841432225065, "percentage": 96.41, "elapsed_time": "13:40:20", "remaining_time": "0:30:32"}
+{"current_steps": 941, "total_steps": 975, "loss": 0.8379, "lr": 2.9631351582350087e-07, "epoch": 4.813299232736573, "percentage": 96.51, "elapsed_time": "13:41:10", "remaining_time": "0:29:40"}
+{"current_steps": 942, "total_steps": 975, "loss": 0.824, "lr": 2.7915964043890186e-07, "epoch": 4.818414322250639, "percentage": 96.62, "elapsed_time": "13:42:02", "remaining_time": "0:28:47"}
+{"current_steps": 943, "total_steps": 975, "loss": 0.838, "lr": 2.6251546961831184e-07, "epoch": 4.823529411764706, "percentage": 96.72, "elapsed_time": "13:42:52", "remaining_time": "0:27:55"}
+{"current_steps": 944, "total_steps": 975, "loss": 0.8465, "lr": 2.4638121694255945e-07, "epoch": 4.828644501278772, "percentage": 96.82, "elapsed_time": "13:43:44", "remaining_time": "0:27:03"}
+{"current_steps": 945, "total_steps": 975, "loss": 0.823, "lr": 2.3075708944910736e-07, "epoch": 4.833759590792839, "percentage": 96.92, "elapsed_time": "13:44:35", "remaining_time": "0:26:10"}
+{"current_steps": 946, "total_steps": 975, "loss": 0.8307, "lr": 2.1564328762941012e-07, "epoch": 4.838874680306906, "percentage": 97.03, "elapsed_time": "13:45:26", "remaining_time": "0:25:18"}
+{"current_steps": 947, "total_steps": 975, "loss": 0.8447, "lr": 2.0104000542634283e-07, "epoch": 4.843989769820972, "percentage": 97.13, "elapsed_time": "13:46:17", "remaining_time": "0:24:25"}
+{"current_steps": 948, "total_steps": 975, "loss": 0.8454, "lr": 1.8694743023170092e-07, "epoch": 4.849104859335038, "percentage": 97.23, "elapsed_time": "13:47:08", "remaining_time": "0:23:33"}
+{"current_steps": 949, "total_steps": 975, "loss": 0.84, "lr": 1.733657428837976e-07, "epoch": 4.854219948849105, "percentage": 97.33, "elapsed_time": "13:47:59", "remaining_time": "0:22:41"}
+{"current_steps": 950, "total_steps": 975, "loss": 0.8474, "lr": 1.6029511766515017e-07, "epoch": 4.859335038363171, "percentage": 97.44, "elapsed_time": "13:48:52", "remaining_time": "0:21:48"}
+{"current_steps": 951, "total_steps": 975, "loss": 0.8307, "lr": 1.4773572230023736e-07, "epoch": 4.864450127877237, "percentage": 97.54, "elapsed_time": "13:49:49", "remaining_time": "0:20:56"}
+{"current_steps": 952, "total_steps": 975, "loss": 0.8491, "lr": 1.3568771795335445e-07, "epoch": 4.869565217391305, "percentage": 97.64, "elapsed_time": "13:50:46", "remaining_time": "0:20:04"}
+{"current_steps": 953, "total_steps": 975, "loss": 0.845, "lr": 1.2415125922653925e-07, "epoch": 4.874680306905371, "percentage": 97.74, "elapsed_time": "13:51:40", "remaining_time": "0:19:11"}
+{"current_steps": 954, "total_steps": 975, "loss": 0.8546, "lr": 1.1312649415758714e-07, "epoch": 4.879795396419437, "percentage": 97.85, "elapsed_time": "13:52:30", "remaining_time": "0:18:19"}
+{"current_steps": 955, "total_steps": 975, "loss": 0.8516, "lr": 1.0261356421815028e-07, "epoch": 4.884910485933504, "percentage": 97.95, "elapsed_time": "13:53:21", "remaining_time": "0:17:27"}
+{"current_steps": 956, "total_steps": 975, "loss": 0.8357, "lr": 9.261260431193464e-08, "epoch": 4.89002557544757, "percentage": 98.05, "elapsed_time": "13:54:12", "remaining_time": "0:16:34"}
+{"current_steps": 957, "total_steps": 975, "loss": 0.8505, "lr": 8.312374277295032e-08, "epoch": 4.8951406649616365, "percentage": 98.15, "elapsed_time": "13:55:03", "remaining_time": "0:15:42"}
+{"current_steps": 958, "total_steps": 975, "loss": 0.8514, "lr": 7.41471013638817e-08, "epoch": 4.900255754475703, "percentage": 98.26, "elapsed_time": "13:55:55", "remaining_time": "0:14:50"}
+{"current_steps": 959, "total_steps": 975, "loss": 0.81, "lr": 6.568279527451538e-08, "epoch": 4.90537084398977, "percentage": 98.36, "elapsed_time": "13:56:45", "remaining_time": "0:13:57"}
+{"current_steps": 960, "total_steps": 975, "loss": 0.8513, "lr": 5.7730933120261345e-08, "epoch": 4.910485933503836, "percentage": 98.46, "elapsed_time": "13:57:36", "remaining_time": "0:13:05"}
+{"current_steps": 961, "total_steps": 975, "loss": 0.8449, "lr": 5.0291616940767454e-08, "epoch": 4.915601023017903, "percentage": 98.56, "elapsed_time": "13:58:26", "remaining_time": "0:12:12"}
+{"current_steps": 962, "total_steps": 975, "loss": 0.8303, "lr": 4.336494219859599e-08, "epoch": 4.920716112531969, "percentage": 98.67, "elapsed_time": "13:59:18", "remaining_time": "0:11:20"}
+{"current_steps": 963, "total_steps": 975, "loss": 0.8629, "lr": 3.695099777802025e-08, "epoch": 4.9258312020460355, "percentage": 98.77, "elapsed_time": "14:00:09", "remaining_time": "0:10:28"}
+{"current_steps": 964, "total_steps": 975, "loss": 0.8439, "lr": 3.1049865983856505e-08, "epoch": 4.930946291560103, "percentage": 98.87, "elapsed_time": "14:01:00", "remaining_time": "0:09:35"}
+{"current_steps": 965, "total_steps": 975, "loss": 0.8576, "lr": 2.5661622540429365e-08, "epoch": 4.936061381074169, "percentage": 98.97, "elapsed_time": "14:01:51", "remaining_time": "0:08:43"}
+{"current_steps": 966, "total_steps": 975, "loss": 0.8511, "lr": 2.0786336590599142e-08, "epoch": 4.9411764705882355, "percentage": 99.08, "elapsed_time": "14:02:46", "remaining_time": "0:07:51"}
+{"current_steps": 967, "total_steps": 975, "loss": 0.8312, "lr": 1.6424070694855965e-08, "epoch": 4.946291560102302, "percentage": 99.18, "elapsed_time": "14:03:45", "remaining_time": "0:06:58"}
+{"current_steps": 968, "total_steps": 975, "loss": 0.8311, "lr": 1.257488083054259e-08, "epoch": 4.951406649616368, "percentage": 99.28, "elapsed_time": "14:04:42", "remaining_time": "0:06:06"}
+{"current_steps": 969, "total_steps": 975, "loss": 0.8294, "lr": 9.238816391108351e-09, "epoch": 4.956521739130435, "percentage": 99.38, "elapsed_time": "14:05:33", "remaining_time": "0:05:14"}
+{"current_steps": 970, "total_steps": 975, "loss": 0.894, "lr": 6.415920185505186e-09, "epoch": 4.961636828644501, "percentage": 99.49, "elapsed_time": "14:06:25", "remaining_time": "0:04:21"}
+{"current_steps": 971, "total_steps": 975, "loss": 0.841, "lr": 4.1062284376147675e-09, "epoch": 4.966751918158568, "percentage": 99.59, "elapsed_time": "14:07:16", "remaining_time": "0:03:29"}
+{"current_steps": 972, "total_steps": 975, "loss": 0.8318, "lr": 2.309770785791088e-09, "epoch": 4.971867007672635, "percentage": 99.69, "elapsed_time": "14:08:07", "remaining_time": "0:02:37"}
+{"current_steps": 973, "total_steps": 975, "loss": 0.8497, "lr": 1.026570282487427e-09, "epoch": 4.976982097186701, "percentage": 99.79, "elapsed_time": "14:08:57", "remaining_time": "0:01:44"}
+{"current_steps": 974, "total_steps": 975, "loss": 0.8306, "lr": 2.566433939454882e-10, "epoch": 4.982097186700767, "percentage": 99.9, "elapsed_time": "14:09:49", "remaining_time": "0:00:52"}
+{"current_steps": 975, "total_steps": 975, "loss": 0.8605, "lr": 0.0, "epoch": 4.987212276214834, "percentage": 100.0, "elapsed_time": "14:10:40", "remaining_time": "0:00:00"}
+{"current_steps": 975, "total_steps": 975, "epoch": 4.987212276214834, "percentage": 100.0, "elapsed_time": "14:12:08", "remaining_time": "0:00:00"}