Training in progress, epoch 1

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b9110dc73ef4180d509694b606ea1e1f790ef28ce4c9290e0c3ae95cbb247c9b
+oid sha256:63369b4c9920c058fa1fb97e31c2a4e134d55aa754748ff280cbb0a73bc7802a
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dfc82023c2ca188997963138670dad0817cd749bc44d518289905946d30023eb
+oid sha256:e02fe7409ac20dba0e6b6d33ed3875c768f6d2f6b80d30c8c798046f3e199e9b
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:40ae4ad8216795ab5c67562512ef23b08030bdddb69f7852eb8ea97bf638bef0
+oid sha256:3e06ff5c3db6321b86182641bb7064b92f72f8b4e3c85a198e3f01180abfa25c
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c0406d31eaec916ea98d7fe4750b83d07dd10379c43d752f685b0ea4550da648
+oid sha256:177ffe80716f86176cc7cdcb7e8d4aa4ff0911ca76edc30a0394868f8a89ee60
 size 1089994880

trainer_log.jsonl

@@ -183,3 +183,176 @@
 {"current_steps": 183, "total_steps": 522, "loss": 0.4328, "lr": 3.2884237380740545e-05, "epoch": 1.0497131931166348, "percentage": 35.06, "elapsed_time": "0:50:12", "remaining_time": "1:33:00"}
 {"current_steps": 184, "total_steps": 522, "loss": 0.4174, "lr": 3.278148265266182e-05, "epoch": 1.0554493307839388, "percentage": 35.25, "elapsed_time": "0:50:29", "remaining_time": "1:32:45"}
 {"current_steps": 185, "total_steps": 522, "loss": 0.3828, "lr": 3.267815442425459e-05, "epoch": 1.0611854684512427, "percentage": 35.44, "elapsed_time": "0:50:44", "remaining_time": "1:32:25"}
+{"current_steps": 186, "total_steps": 522, "loss": 0.4549, "lr": 3.25742573318177e-05, "epoch": 1.066921606118547, "percentage": 35.63, "elapsed_time": "0:50:57", "remaining_time": "1:32:02"}
+{"current_steps": 187, "total_steps": 522, "loss": 0.4203, "lr": 3.246979603717467e-05, "epoch": 1.0726577437858509, "percentage": 35.82, "elapsed_time": "0:51:15", "remaining_time": "1:31:48"}
+{"current_steps": 188, "total_steps": 522, "loss": 0.4675, "lr": 3.236477522746459e-05, "epoch": 1.0783938814531548, "percentage": 36.02, "elapsed_time": "0:51:34", "remaining_time": "1:31:36"}
+{"current_steps": 189, "total_steps": 522, "loss": 0.3542, "lr": 3.2259199614931745e-05, "epoch": 1.0841300191204588, "percentage": 36.21, "elapsed_time": "0:51:50", "remaining_time": "1:31:20"}
+{"current_steps": 190, "total_steps": 522, "loss": 0.4785, "lr": 3.215307393671426e-05, "epoch": 1.089866156787763, "percentage": 36.4, "elapsed_time": "0:52:11", "remaining_time": "1:31:11"}
+{"current_steps": 191, "total_steps": 522, "loss": 0.3886, "lr": 3.204640295463146e-05, "epoch": 1.095602294455067, "percentage": 36.59, "elapsed_time": "0:52:25", "remaining_time": "1:30:50"}
+{"current_steps": 192, "total_steps": 522, "loss": 0.3842, "lr": 3.193919145497028e-05, "epoch": 1.101338432122371, "percentage": 36.78, "elapsed_time": "0:52:40", "remaining_time": "1:30:32"}
+{"current_steps": 193, "total_steps": 522, "loss": 0.423, "lr": 3.1831444248270455e-05, "epoch": 1.107074569789675, "percentage": 36.97, "elapsed_time": "0:52:54", "remaining_time": "1:30:10"}
+{"current_steps": 194, "total_steps": 522, "loss": 0.4239, "lr": 3.17231661691087e-05, "epoch": 1.1128107074569789, "percentage": 37.16, "elapsed_time": "0:53:09", "remaining_time": "1:29:53"}
+{"current_steps": 195, "total_steps": 522, "loss": 0.425, "lr": 3.161436207588178e-05, "epoch": 1.118546845124283, "percentage": 37.36, "elapsed_time": "0:53:24", "remaining_time": "1:29:34"}
+{"current_steps": 196, "total_steps": 522, "loss": 0.3557, "lr": 3.1505036850588516e-05, "epoch": 1.124282982791587, "percentage": 37.55, "elapsed_time": "0:53:37", "remaining_time": "1:29:10"}
+{"current_steps": 197, "total_steps": 522, "loss": 0.4791, "lr": 3.139519539861074e-05, "epoch": 1.130019120458891, "percentage": 37.74, "elapsed_time": "0:53:54", "remaining_time": "1:28:56"}
+{"current_steps": 198, "total_steps": 522, "loss": 0.4214, "lr": 3.128484264849314e-05, "epoch": 1.135755258126195, "percentage": 37.93, "elapsed_time": "0:54:11", "remaining_time": "1:28:41"}
+{"current_steps": 199, "total_steps": 522, "loss": 0.3698, "lr": 3.1173983551722205e-05, "epoch": 1.1414913957934991, "percentage": 38.12, "elapsed_time": "0:54:23", "remaining_time": "1:28:16"}
+{"current_steps": 200, "total_steps": 522, "loss": 0.3882, "lr": 3.106262308250399e-05, "epoch": 1.147227533460803, "percentage": 38.31, "elapsed_time": "0:54:35", "remaining_time": "1:27:52"}
+{"current_steps": 201, "total_steps": 522, "loss": 0.3937, "lr": 3.0950766237540946e-05, "epoch": 1.152963671128107, "percentage": 38.51, "elapsed_time": "0:54:52", "remaining_time": "1:27:38"}
+{"current_steps": 202, "total_steps": 522, "loss": 0.4585, "lr": 3.083841803580771e-05, "epoch": 1.158699808795411, "percentage": 38.7, "elapsed_time": "0:55:05", "remaining_time": "1:27:16"}
+{"current_steps": 203, "total_steps": 522, "loss": 0.3764, "lr": 3.072558351832592e-05, "epoch": 1.1644359464627152, "percentage": 38.89, "elapsed_time": "0:55:20", "remaining_time": "1:26:57"}
+{"current_steps": 204, "total_steps": 522, "loss": 0.4736, "lr": 3.0612267747938015e-05, "epoch": 1.1701720841300192, "percentage": 39.08, "elapsed_time": "0:55:37", "remaining_time": "1:26:42"}
+{"current_steps": 205, "total_steps": 522, "loss": 0.4208, "lr": 3.0498475809080082e-05, "epoch": 1.1759082217973231, "percentage": 39.27, "elapsed_time": "0:55:53", "remaining_time": "1:26:25"}
+{"current_steps": 206, "total_steps": 522, "loss": 0.4024, "lr": 3.03842128075537e-05, "epoch": 1.181644359464627, "percentage": 39.46, "elapsed_time": "0:56:08", "remaining_time": "1:26:07"}
+{"current_steps": 207, "total_steps": 522, "loss": 0.4055, "lr": 3.026948387029684e-05, "epoch": 1.1873804971319313, "percentage": 39.66, "elapsed_time": "0:56:22", "remaining_time": "1:25:47"}
+{"current_steps": 208, "total_steps": 522, "loss": 0.4051, "lr": 3.0154294145153865e-05, "epoch": 1.1931166347992352, "percentage": 39.85, "elapsed_time": "0:56:38", "remaining_time": "1:25:30"}
+{"current_steps": 209, "total_steps": 522, "loss": 0.41, "lr": 3.003864880064449e-05, "epoch": 1.1988527724665392, "percentage": 40.04, "elapsed_time": "0:56:54", "remaining_time": "1:25:13"}
+{"current_steps": 210, "total_steps": 522, "loss": 0.4053, "lr": 2.9922553025731907e-05, "epoch": 1.2045889101338432, "percentage": 40.23, "elapsed_time": "0:57:12", "remaining_time": "1:24:59"}
+{"current_steps": 211, "total_steps": 522, "loss": 0.4272, "lr": 2.9806012029589966e-05, "epoch": 1.2103250478011471, "percentage": 40.42, "elapsed_time": "0:57:29", "remaining_time": "1:24:44"}
+{"current_steps": 212, "total_steps": 522, "loss": 0.4182, "lr": 2.9689031041369406e-05, "epoch": 1.2160611854684513, "percentage": 40.61, "elapsed_time": "0:57:43", "remaining_time": "1:24:25"}
+{"current_steps": 213, "total_steps": 522, "loss": 0.3871, "lr": 2.9571615309963255e-05, "epoch": 1.2217973231357553, "percentage": 40.8, "elapsed_time": "0:57:54", "remaining_time": "1:24:00"}
+{"current_steps": 214, "total_steps": 522, "loss": 0.39, "lr": 2.9453770103771308e-05, "epoch": 1.2275334608030593, "percentage": 41.0, "elapsed_time": "0:58:08", "remaining_time": "1:23:40"}
+{"current_steps": 215, "total_steps": 522, "loss": 0.435, "lr": 2.9335500710463725e-05, "epoch": 1.2332695984703632, "percentage": 41.19, "elapsed_time": "0:58:24", "remaining_time": "1:23:24"}
+{"current_steps": 216, "total_steps": 522, "loss": 0.4313, "lr": 2.9216812436743788e-05, "epoch": 1.2390057361376674, "percentage": 41.38, "elapsed_time": "0:58:39", "remaining_time": "1:23:06"}
+{"current_steps": 217, "total_steps": 522, "loss": 0.3601, "lr": 2.9097710608109776e-05, "epoch": 1.2447418738049714, "percentage": 41.57, "elapsed_time": "0:58:55", "remaining_time": "1:22:49"}
+{"current_steps": 218, "total_steps": 522, "loss": 0.3711, "lr": 2.8978200568616036e-05, "epoch": 1.2504780114722753, "percentage": 41.76, "elapsed_time": "0:59:07", "remaining_time": "1:22:26"}
+{"current_steps": 219, "total_steps": 522, "loss": 0.4385, "lr": 2.885828768063317e-05, "epoch": 1.2562141491395793, "percentage": 41.95, "elapsed_time": "0:59:21", "remaining_time": "1:22:07"}
+{"current_steps": 220, "total_steps": 522, "loss": 0.4318, "lr": 2.8737977324607466e-05, "epoch": 1.2619502868068833, "percentage": 42.15, "elapsed_time": "0:59:38", "remaining_time": "1:21:52"}
+{"current_steps": 221, "total_steps": 522, "loss": 0.4165, "lr": 2.861727489881941e-05, "epoch": 1.2676864244741874, "percentage": 42.34, "elapsed_time": "0:59:55", "remaining_time": "1:21:36"}
+{"current_steps": 222, "total_steps": 522, "loss": 0.4346, "lr": 2.8496185819141547e-05, "epoch": 1.2734225621414914, "percentage": 42.53, "elapsed_time": "1:00:12", "remaining_time": "1:21:21"}
+{"current_steps": 223, "total_steps": 522, "loss": 0.4488, "lr": 2.837471551879543e-05, "epoch": 1.2791586998087954, "percentage": 42.72, "elapsed_time": "1:00:28", "remaining_time": "1:21:04"}
+{"current_steps": 224, "total_steps": 522, "loss": 0.4172, "lr": 2.825286944810783e-05, "epoch": 1.2848948374760996, "percentage": 42.91, "elapsed_time": "1:00:43", "remaining_time": "1:20:47"}
+{"current_steps": 225, "total_steps": 522, "loss": 0.3899, "lr": 2.8130653074266188e-05, "epoch": 1.2906309751434035, "percentage": 43.1, "elapsed_time": "1:00:57", "remaining_time": "1:20:28"}
+{"current_steps": 226, "total_steps": 522, "loss": 0.4119, "lr": 2.8008071881073342e-05, "epoch": 1.2963671128107075, "percentage": 43.3, "elapsed_time": "1:01:11", "remaining_time": "1:20:09"}
+{"current_steps": 227, "total_steps": 522, "loss": 0.4167, "lr": 2.78851313687014e-05, "epoch": 1.3021032504780115, "percentage": 43.49, "elapsed_time": "1:01:24", "remaining_time": "1:19:48"}
+{"current_steps": 228, "total_steps": 522, "loss": 0.4428, "lr": 2.7761837053444994e-05, "epoch": 1.3078393881453154, "percentage": 43.68, "elapsed_time": "1:01:44", "remaining_time": "1:19:36"}
+{"current_steps": 229, "total_steps": 522, "loss": 0.4623, "lr": 2.7638194467473776e-05, "epoch": 1.3135755258126194, "percentage": 43.87, "elapsed_time": "1:02:01", "remaining_time": "1:19:21"}
+{"current_steps": 230, "total_steps": 522, "loss": 0.3855, "lr": 2.7514209158584164e-05, "epoch": 1.3193116634799236, "percentage": 44.06, "elapsed_time": "1:02:16", "remaining_time": "1:19:03"}
+{"current_steps": 231, "total_steps": 522, "loss": 0.3704, "lr": 2.7389886689950428e-05, "epoch": 1.3250478011472275, "percentage": 44.25, "elapsed_time": "1:02:31", "remaining_time": "1:18:46"}
+{"current_steps": 232, "total_steps": 522, "loss": 0.4346, "lr": 2.7265232639875043e-05, "epoch": 1.3307839388145315, "percentage": 44.44, "elapsed_time": "1:02:50", "remaining_time": "1:18:33"}
+{"current_steps": 233, "total_steps": 522, "loss": 0.4339, "lr": 2.7140252601538466e-05, "epoch": 1.3365200764818357, "percentage": 44.64, "elapsed_time": "1:03:07", "remaining_time": "1:18:17"}
+{"current_steps": 234, "total_steps": 522, "loss": 0.4307, "lr": 2.7014952182748083e-05, "epoch": 1.3422562141491396, "percentage": 44.83, "elapsed_time": "1:03:26", "remaining_time": "1:18:04"}
+{"current_steps": 235, "total_steps": 522, "loss": 0.4091, "lr": 2.688933700568666e-05, "epoch": 1.3479923518164436, "percentage": 45.02, "elapsed_time": "1:03:43", "remaining_time": "1:17:48"}
+{"current_steps": 236, "total_steps": 522, "loss": 0.4544, "lr": 2.676341270666003e-05, "epoch": 1.3537284894837476, "percentage": 45.21, "elapsed_time": "1:04:01", "remaining_time": "1:17:34"}
+{"current_steps": 237, "total_steps": 522, "loss": 0.3461, "lr": 2.663718493584422e-05, "epoch": 1.3594646271510515, "percentage": 45.4, "elapsed_time": "1:04:15", "remaining_time": "1:17:16"}
+{"current_steps": 238, "total_steps": 522, "loss": 0.4336, "lr": 2.6510659357031917e-05, "epoch": 1.3652007648183555, "percentage": 45.59, "elapsed_time": "1:04:33", "remaining_time": "1:17:02"}
+{"current_steps": 239, "total_steps": 522, "loss": 0.38, "lr": 2.6383841647378356e-05, "epoch": 1.3709369024856597, "percentage": 45.79, "elapsed_time": "1:04:50", "remaining_time": "1:16:46"}
+{"current_steps": 240, "total_steps": 522, "loss": 0.4709, "lr": 2.6256737497146544e-05, "epoch": 1.3766730401529637, "percentage": 45.98, "elapsed_time": "1:05:05", "remaining_time": "1:16:29"}
+{"current_steps": 241, "total_steps": 522, "loss": 0.4105, "lr": 2.6129352609452003e-05, "epoch": 1.3824091778202676, "percentage": 46.17, "elapsed_time": "1:05:22", "remaining_time": "1:16:13"}
+{"current_steps": 242, "total_steps": 522, "loss": 0.3782, "lr": 2.600169270000682e-05, "epoch": 1.3881453154875718, "percentage": 46.36, "elapsed_time": "1:05:38", "remaining_time": "1:15:56"}
+{"current_steps": 243, "total_steps": 522, "loss": 0.3954, "lr": 2.5873763496863203e-05, "epoch": 1.3938814531548758, "percentage": 46.55, "elapsed_time": "1:05:52", "remaining_time": "1:15:38"}
+{"current_steps": 244, "total_steps": 522, "loss": 0.458, "lr": 2.574557074015648e-05, "epoch": 1.3996175908221797, "percentage": 46.74, "elapsed_time": "1:06:10", "remaining_time": "1:15:24"}
+{"current_steps": 245, "total_steps": 522, "loss": 0.438, "lr": 2.561712018184752e-05, "epoch": 1.4053537284894837, "percentage": 46.93, "elapsed_time": "1:06:25", "remaining_time": "1:15:05"}
+{"current_steps": 246, "total_steps": 522, "loss": 0.4171, "lr": 2.5488417585464648e-05, "epoch": 1.4110898661567877, "percentage": 47.13, "elapsed_time": "1:06:43", "remaining_time": "1:14:51"}
+{"current_steps": 247, "total_steps": 522, "loss": 0.4265, "lr": 2.535946872584506e-05, "epoch": 1.4168260038240919, "percentage": 47.32, "elapsed_time": "1:07:01", "remaining_time": "1:14:37"}
+{"current_steps": 248, "total_steps": 522, "loss": 0.39, "lr": 2.523027938887567e-05, "epoch": 1.4225621414913958, "percentage": 47.51, "elapsed_time": "1:07:19", "remaining_time": "1:14:23"}
+{"current_steps": 249, "total_steps": 522, "loss": 0.4255, "lr": 2.5100855371233533e-05, "epoch": 1.4282982791586998, "percentage": 47.7, "elapsed_time": "1:07:36", "remaining_time": "1:14:07"}
+{"current_steps": 250, "total_steps": 522, "loss": 0.3925, "lr": 2.4971202480125737e-05, "epoch": 1.4340344168260037, "percentage": 47.89, "elapsed_time": "1:07:55", "remaining_time": "1:13:54"}
+{"current_steps": 251, "total_steps": 522, "loss": 0.4283, "lr": 2.4841326533028838e-05, "epoch": 1.439770554493308, "percentage": 48.08, "elapsed_time": "1:08:13", "remaining_time": "1:13:39"}
+{"current_steps": 252, "total_steps": 522, "loss": 0.429, "lr": 2.4711233357427817e-05, "epoch": 1.445506692160612, "percentage": 48.28, "elapsed_time": "1:08:30", "remaining_time": "1:13:23"}
+{"current_steps": 253, "total_steps": 522, "loss": 0.3684, "lr": 2.458092879055464e-05, "epoch": 1.4512428298279159, "percentage": 48.47, "elapsed_time": "1:08:44", "remaining_time": "1:13:05"}
+{"current_steps": 254, "total_steps": 522, "loss": 0.4141, "lr": 2.445041867912629e-05, "epoch": 1.4569789674952198, "percentage": 48.66, "elapsed_time": "1:08:59", "remaining_time": "1:12:47"}
+{"current_steps": 255, "total_steps": 522, "loss": 0.428, "lr": 2.431970887908249e-05, "epoch": 1.4627151051625238, "percentage": 48.85, "elapsed_time": "1:09:16", "remaining_time": "1:12:31"}
+{"current_steps": 256, "total_steps": 522, "loss": 0.422, "lr": 2.4188805255322895e-05, "epoch": 1.468451242829828, "percentage": 49.04, "elapsed_time": "1:09:35", "remaining_time": "1:12:18"}
+{"current_steps": 257, "total_steps": 522, "loss": 0.4407, "lr": 2.405771368144395e-05, "epoch": 1.474187380497132, "percentage": 49.23, "elapsed_time": "1:09:53", "remaining_time": "1:12:04"}
+{"current_steps": 258, "total_steps": 522, "loss": 0.4004, "lr": 2.3926440039475382e-05, "epoch": 1.479923518164436, "percentage": 49.43, "elapsed_time": "1:10:09", "remaining_time": "1:11:47"}
+{"current_steps": 259, "total_steps": 522, "loss": 0.433, "lr": 2.3794990219616217e-05, "epoch": 1.48565965583174, "percentage": 49.62, "elapsed_time": "1:10:26", "remaining_time": "1:11:31"}
+{"current_steps": 260, "total_steps": 522, "loss": 0.4281, "lr": 2.3663370119970527e-05, "epoch": 1.491395793499044, "percentage": 49.81, "elapsed_time": "1:10:41", "remaining_time": "1:11:13"}
+{"current_steps": 261, "total_steps": 522, "loss": 0.3873, "lr": 2.353158564628277e-05, "epoch": 1.497131931166348, "percentage": 50.0, "elapsed_time": "1:10:52", "remaining_time": "1:10:52"}
+{"current_steps": 262, "total_steps": 522, "loss": 0.3906, "lr": 2.339964271167282e-05, "epoch": 1.502868068833652, "percentage": 50.19, "elapsed_time": "1:11:05", "remaining_time": "1:10:32"}
+{"current_steps": 263, "total_steps": 522, "loss": 0.4153, "lr": 2.3267547236370604e-05, "epoch": 1.508604206500956, "percentage": 50.38, "elapsed_time": "1:11:20", "remaining_time": "1:10:14"}
+{"current_steps": 264, "total_steps": 522, "loss": 0.4243, "lr": 2.3135305147450525e-05, "epoch": 1.51434034416826, "percentage": 50.57, "elapsed_time": "1:11:36", "remaining_time": "1:09:59"}
+{"current_steps": 265, "total_steps": 522, "loss": 0.379, "lr": 2.3002922378565468e-05, "epoch": 1.520076481835564, "percentage": 50.77, "elapsed_time": "1:11:53", "remaining_time": "1:09:43"}
+{"current_steps": 266, "total_steps": 522, "loss": 0.3992, "lr": 2.2870404869680573e-05, "epoch": 1.525812619502868, "percentage": 50.96, "elapsed_time": "1:12:08", "remaining_time": "1:09:25"}
+{"current_steps": 267, "total_steps": 522, "loss": 0.4029, "lr": 2.273775856680672e-05, "epoch": 1.5315487571701722, "percentage": 51.15, "elapsed_time": "1:12:25", "remaining_time": "1:09:10"}
+{"current_steps": 268, "total_steps": 522, "loss": 0.3964, "lr": 2.2604989421733718e-05, "epoch": 1.5372848948374762, "percentage": 51.34, "elapsed_time": "1:12:41", "remaining_time": "1:08:54"}
+{"current_steps": 269, "total_steps": 522, "loss": 0.4573, "lr": 2.2472103391763268e-05, "epoch": 1.5430210325047802, "percentage": 51.53, "elapsed_time": "1:13:00", "remaining_time": "1:08:39"}
+{"current_steps": 270, "total_steps": 522, "loss": 0.4331, "lr": 2.2339106439441656e-05, "epoch": 1.5487571701720841, "percentage": 51.72, "elapsed_time": "1:13:13", "remaining_time": "1:08:20"}
+{"current_steps": 271, "total_steps": 522, "loss": 0.4301, "lr": 2.2206004532292192e-05, "epoch": 1.554493307839388, "percentage": 51.92, "elapsed_time": "1:13:30", "remaining_time": "1:08:04"}
+{"current_steps": 272, "total_steps": 522, "loss": 0.4029, "lr": 2.20728036425475e-05, "epoch": 1.560229445506692, "percentage": 52.11, "elapsed_time": "1:13:47", "remaining_time": "1:07:49"}
+{"current_steps": 273, "total_steps": 522, "loss": 0.4204, "lr": 2.1939509746881486e-05, "epoch": 1.565965583173996, "percentage": 52.3, "elapsed_time": "1:14:04", "remaining_time": "1:07:33"}
+{"current_steps": 274, "total_steps": 522, "loss": 0.4556, "lr": 2.1806128826141223e-05, "epoch": 1.5717017208413002, "percentage": 52.49, "elapsed_time": "1:14:20", "remaining_time": "1:07:16"}
+{"current_steps": 275, "total_steps": 522, "loss": 0.3854, "lr": 2.1672666865078544e-05, "epoch": 1.5774378585086042, "percentage": 52.68, "elapsed_time": "1:14:35", "remaining_time": "1:06:59"}
+{"current_steps": 276, "total_steps": 522, "loss": 0.4304, "lr": 2.153912985208154e-05, "epoch": 1.5831739961759084, "percentage": 52.87, "elapsed_time": "1:14:55", "remaining_time": "1:06:47"}
+{"current_steps": 277, "total_steps": 522, "loss": 0.413, "lr": 2.140552377890586e-05, "epoch": 1.5889101338432123, "percentage": 53.07, "elapsed_time": "1:15:13", "remaining_time": "1:06:31"}
+{"current_steps": 278, "total_steps": 522, "loss": 0.4033, "lr": 2.1271854640405856e-05, "epoch": 1.5946462715105163, "percentage": 53.26, "elapsed_time": "1:15:29", "remaining_time": "1:06:15"}
+{"current_steps": 279, "total_steps": 522, "loss": 0.3809, "lr": 2.1138128434265583e-05, "epoch": 1.6003824091778203, "percentage": 53.45, "elapsed_time": "1:15:45", "remaining_time": "1:05:59"}
+{"current_steps": 280, "total_steps": 522, "loss": 0.4356, "lr": 2.1004351160729714e-05, "epoch": 1.6061185468451242, "percentage": 53.64, "elapsed_time": "1:16:02", "remaining_time": "1:05:43"}
+{"current_steps": 281, "total_steps": 522, "loss": 0.4071, "lr": 2.0870528822334296e-05, "epoch": 1.6118546845124282, "percentage": 53.83, "elapsed_time": "1:16:17", "remaining_time": "1:05:25"}
+{"current_steps": 282, "total_steps": 522, "loss": 0.4566, "lr": 2.073666742363742e-05, "epoch": 1.6175908221797322, "percentage": 54.02, "elapsed_time": "1:16:33", "remaining_time": "1:05:09"}
+{"current_steps": 283, "total_steps": 522, "loss": 0.4084, "lr": 2.0602772970949777e-05, "epoch": 1.6233269598470363, "percentage": 54.21, "elapsed_time": "1:16:51", "remaining_time": "1:04:54"}
+{"current_steps": 284, "total_steps": 522, "loss": 0.4276, "lr": 2.0468851472065213e-05, "epoch": 1.6290630975143403, "percentage": 54.41, "elapsed_time": "1:17:11", "remaining_time": "1:04:41"}
+{"current_steps": 285, "total_steps": 522, "loss": 0.3797, "lr": 2.03349089359911e-05, "epoch": 1.6347992351816445, "percentage": 54.6, "elapsed_time": "1:17:26", "remaining_time": "1:04:23"}
+{"current_steps": 286, "total_steps": 522, "loss": 0.4344, "lr": 2.020095137267876e-05, "epoch": 1.6405353728489485, "percentage": 54.79, "elapsed_time": "1:17:39", "remaining_time": "1:04:04"}
+{"current_steps": 287, "total_steps": 522, "loss": 0.4187, "lr": 2.0066984792753757e-05, "epoch": 1.6462715105162524, "percentage": 54.98, "elapsed_time": "1:17:59", "remaining_time": "1:03:51"}
+{"current_steps": 288, "total_steps": 522, "loss": 0.4036, "lr": 1.993301520724625e-05, "epoch": 1.6520076481835564, "percentage": 55.17, "elapsed_time": "1:18:13", "remaining_time": "1:03:33"}
+{"current_steps": 289, "total_steps": 522, "loss": 0.4141, "lr": 1.9799048627321247e-05, "epoch": 1.6577437858508604, "percentage": 55.36, "elapsed_time": "1:18:30", "remaining_time": "1:03:17"}
+{"current_steps": 290, "total_steps": 522, "loss": 0.4098, "lr": 1.966509106400891e-05, "epoch": 1.6634799235181643, "percentage": 55.56, "elapsed_time": "1:18:46", "remaining_time": "1:03:01"}
+{"current_steps": 291, "total_steps": 522, "loss": 0.372, "lr": 1.9531148527934794e-05, "epoch": 1.6692160611854685, "percentage": 55.75, "elapsed_time": "1:19:00", "remaining_time": "1:02:42"}
+{"current_steps": 292, "total_steps": 522, "loss": 0.3977, "lr": 1.9397227029050233e-05, "epoch": 1.6749521988527725, "percentage": 55.94, "elapsed_time": "1:19:16", "remaining_time": "1:02:26"}
+{"current_steps": 293, "total_steps": 522, "loss": 0.4654, "lr": 1.9263332576362586e-05, "epoch": 1.6806883365200764, "percentage": 56.13, "elapsed_time": "1:19:32", "remaining_time": "1:02:10"}
+{"current_steps": 294, "total_steps": 522, "loss": 0.5054, "lr": 1.9129471177665714e-05, "epoch": 1.6864244741873806, "percentage": 56.32, "elapsed_time": "1:19:48", "remaining_time": "1:01:53"}
+{"current_steps": 295, "total_steps": 522, "loss": 0.3574, "lr": 1.899564883927029e-05, "epoch": 1.6921606118546846, "percentage": 56.51, "elapsed_time": "1:20:01", "remaining_time": "1:01:34"}
+{"current_steps": 296, "total_steps": 522, "loss": 0.408, "lr": 1.8861871565734427e-05, "epoch": 1.6978967495219885, "percentage": 56.7, "elapsed_time": "1:20:18", "remaining_time": "1:01:19"}
+{"current_steps": 297, "total_steps": 522, "loss": 0.414, "lr": 1.8728145359594147e-05, "epoch": 1.7036328871892925, "percentage": 56.9, "elapsed_time": "1:20:36", "remaining_time": "1:01:03"}
+{"current_steps": 298, "total_steps": 522, "loss": 0.4452, "lr": 1.859447622109415e-05, "epoch": 1.7093690248565965, "percentage": 57.09, "elapsed_time": "1:20:54", "remaining_time": "1:00:49"}
+{"current_steps": 299, "total_steps": 522, "loss": 0.4098, "lr": 1.8460870147918464e-05, "epoch": 1.7151051625239004, "percentage": 57.28, "elapsed_time": "1:21:11", "remaining_time": "1:00:33"}
+{"current_steps": 300, "total_steps": 522, "loss": 0.3665, "lr": 1.832733313492147e-05, "epoch": 1.7208413001912046, "percentage": 57.47, "elapsed_time": "1:21:27", "remaining_time": "1:00:16"}
+{"current_steps": 301, "total_steps": 522, "loss": 0.4017, "lr": 1.8193871173858784e-05, "epoch": 1.7265774378585086, "percentage": 57.66, "elapsed_time": "1:21:38", "remaining_time": "0:59:56"}
+{"current_steps": 302, "total_steps": 522, "loss": 0.457, "lr": 1.8060490253118524e-05, "epoch": 1.7323135755258128, "percentage": 57.85, "elapsed_time": "1:21:52", "remaining_time": "0:59:38"}
+{"current_steps": 303, "total_steps": 522, "loss": 0.3524, "lr": 1.7927196357452507e-05, "epoch": 1.7380497131931167, "percentage": 58.05, "elapsed_time": "1:22:11", "remaining_time": "0:59:24"}
+{"current_steps": 304, "total_steps": 522, "loss": 0.4125, "lr": 1.7793995467707808e-05, "epoch": 1.7437858508604207, "percentage": 58.24, "elapsed_time": "1:22:26", "remaining_time": "0:59:07"}
+{"current_steps": 305, "total_steps": 522, "loss": 0.4564, "lr": 1.766089356055835e-05, "epoch": 1.7495219885277247, "percentage": 58.43, "elapsed_time": "1:22:43", "remaining_time": "0:58:51"}
+{"current_steps": 306, "total_steps": 522, "loss": 0.3906, "lr": 1.7527896608236735e-05, "epoch": 1.7552581261950286, "percentage": 58.62, "elapsed_time": "1:23:00", "remaining_time": "0:58:35"}
+{"current_steps": 307, "total_steps": 522, "loss": 0.4225, "lr": 1.7395010578266292e-05, "epoch": 1.7609942638623326, "percentage": 58.81, "elapsed_time": "1:23:20", "remaining_time": "0:58:21"}
+{"current_steps": 308, "total_steps": 522, "loss": 0.3477, "lr": 1.7262241433193284e-05, "epoch": 1.7667304015296366, "percentage": 59.0, "elapsed_time": "1:23:40", "remaining_time": "0:58:07"}
+{"current_steps": 309, "total_steps": 522, "loss": 0.3823, "lr": 1.712959513031943e-05, "epoch": 1.7724665391969407, "percentage": 59.2, "elapsed_time": "1:23:57", "remaining_time": "0:57:52"}
+{"current_steps": 310, "total_steps": 522, "loss": 0.4066, "lr": 1.6997077621434535e-05, "epoch": 1.7782026768642447, "percentage": 59.39, "elapsed_time": "1:24:14", "remaining_time": "0:57:36"}
+{"current_steps": 311, "total_steps": 522, "loss": 0.4268, "lr": 1.6864694852549478e-05, "epoch": 1.783938814531549, "percentage": 59.58, "elapsed_time": "1:24:28", "remaining_time": "0:57:18"}
+{"current_steps": 312, "total_steps": 522, "loss": 0.4326, "lr": 1.6732452763629396e-05, "epoch": 1.7896749521988529, "percentage": 59.77, "elapsed_time": "1:24:44", "remaining_time": "0:57:02"}
+{"current_steps": 313, "total_steps": 522, "loss": 0.379, "lr": 1.6600357288327186e-05, "epoch": 1.7954110898661568, "percentage": 59.96, "elapsed_time": "1:24:53", "remaining_time": "0:56:41"}
+{"current_steps": 314, "total_steps": 522, "loss": 0.432, "lr": 1.646841435371723e-05, "epoch": 1.8011472275334608, "percentage": 60.15, "elapsed_time": "1:25:09", "remaining_time": "0:56:24"}
+{"current_steps": 315, "total_steps": 522, "loss": 0.4023, "lr": 1.6336629880029483e-05, "epoch": 1.8068833652007648, "percentage": 60.34, "elapsed_time": "1:25:25", "remaining_time": "0:56:07"}
+{"current_steps": 316, "total_steps": 522, "loss": 0.4308, "lr": 1.6205009780383783e-05, "epoch": 1.8126195028680687, "percentage": 60.54, "elapsed_time": "1:25:41", "remaining_time": "0:55:51"}
+{"current_steps": 317, "total_steps": 522, "loss": 0.3976, "lr": 1.6073559960524624e-05, "epoch": 1.8183556405353727, "percentage": 60.73, "elapsed_time": "1:25:58", "remaining_time": "0:55:36"}
+{"current_steps": 318, "total_steps": 522, "loss": 0.3835, "lr": 1.594228631855605e-05, "epoch": 1.8240917782026769, "percentage": 60.92, "elapsed_time": "1:26:13", "remaining_time": "0:55:18"}
+{"current_steps": 319, "total_steps": 522, "loss": 0.4129, "lr": 1.5811194744677116e-05, "epoch": 1.8298279158699808, "percentage": 61.11, "elapsed_time": "1:26:28", "remaining_time": "0:55:01"}
+{"current_steps": 320, "total_steps": 522, "loss": 0.36, "lr": 1.5680291120917512e-05, "epoch": 1.835564053537285, "percentage": 61.3, "elapsed_time": "1:26:42", "remaining_time": "0:54:44"}
+{"current_steps": 321, "total_steps": 522, "loss": 0.3892, "lr": 1.5549581320873715e-05, "epoch": 1.841300191204589, "percentage": 61.49, "elapsed_time": "1:26:59", "remaining_time": "0:54:28"}
+{"current_steps": 322, "total_steps": 522, "loss": 0.3865, "lr": 1.541907120944537e-05, "epoch": 1.847036328871893, "percentage": 61.69, "elapsed_time": "1:27:15", "remaining_time": "0:54:11"}
+{"current_steps": 323, "total_steps": 522, "loss": 0.4356, "lr": 1.528876664257219e-05, "epoch": 1.852772466539197, "percentage": 61.88, "elapsed_time": "1:27:28", "remaining_time": "0:53:53"}
+{"current_steps": 324, "total_steps": 522, "loss": 0.4257, "lr": 1.5158673466971168e-05, "epoch": 1.8585086042065009, "percentage": 62.07, "elapsed_time": "1:27:46", "remaining_time": "0:53:38"}
+{"current_steps": 325, "total_steps": 522, "loss": 0.3947, "lr": 1.502879751987427e-05, "epoch": 1.8642447418738048, "percentage": 62.26, "elapsed_time": "1:28:00", "remaining_time": "0:53:20"}
+{"current_steps": 326, "total_steps": 522, "loss": 0.4197, "lr": 1.4899144628766473e-05, "epoch": 1.869980879541109, "percentage": 62.45, "elapsed_time": "1:28:18", "remaining_time": "0:53:05"}
+{"current_steps": 327, "total_steps": 522, "loss": 0.3862, "lr": 1.4769720611124342e-05, "epoch": 1.875717017208413, "percentage": 62.64, "elapsed_time": "1:28:33", "remaining_time": "0:52:48"}
+{"current_steps": 328, "total_steps": 522, "loss": 0.4547, "lr": 1.4640531274154946e-05, "epoch": 1.8814531548757172, "percentage": 62.84, "elapsed_time": "1:28:48", "remaining_time": "0:52:31"}
+{"current_steps": 329, "total_steps": 522, "loss": 0.372, "lr": 1.4511582414535359e-05, "epoch": 1.8871892925430211, "percentage": 63.03, "elapsed_time": "1:29:05", "remaining_time": "0:52:15"}
+{"current_steps": 330, "total_steps": 522, "loss": 0.391, "lr": 1.4382879818152486e-05, "epoch": 1.892925430210325, "percentage": 63.22, "elapsed_time": "1:29:20", "remaining_time": "0:51:58"}
+{"current_steps": 331, "total_steps": 522, "loss": 0.3807, "lr": 1.425442925984353e-05, "epoch": 1.898661567877629, "percentage": 63.41, "elapsed_time": "1:29:38", "remaining_time": "0:51:43"}
+{"current_steps": 332, "total_steps": 522, "loss": 0.376, "lr": 1.41262365031368e-05, "epoch": 1.904397705544933, "percentage": 63.6, "elapsed_time": "1:29:55", "remaining_time": "0:51:27"}
+{"current_steps": 333, "total_steps": 522, "loss": 0.3761, "lr": 1.399830729999319e-05, "epoch": 1.910133843212237, "percentage": 63.79, "elapsed_time": "1:30:09", "remaining_time": "0:51:10"}
+{"current_steps": 334, "total_steps": 522, "loss": 0.3989, "lr": 1.3870647390548004e-05, "epoch": 1.915869980879541, "percentage": 63.98, "elapsed_time": "1:30:24", "remaining_time": "0:50:53"}
+{"current_steps": 335, "total_steps": 522, "loss": 0.4614, "lr": 1.3743262502853458e-05, "epoch": 1.9216061185468452, "percentage": 64.18, "elapsed_time": "1:30:42", "remaining_time": "0:50:37"}
+{"current_steps": 336, "total_steps": 522, "loss": 0.3545, "lr": 1.3616158352621653e-05, "epoch": 1.9273422562141491, "percentage": 64.37, "elapsed_time": "1:30:57", "remaining_time": "0:50:21"}
+{"current_steps": 337, "total_steps": 522, "loss": 0.3725, "lr": 1.3489340642968088e-05, "epoch": 1.9330783938814533, "percentage": 64.56, "elapsed_time": "1:31:12", "remaining_time": "0:50:04"}
+{"current_steps": 338, "total_steps": 522, "loss": 0.4091, "lr": 1.336281506415579e-05, "epoch": 1.9388145315487573, "percentage": 64.75, "elapsed_time": "1:31:30", "remaining_time": "0:49:48"}
+{"current_steps": 339, "total_steps": 522, "loss": 0.4097, "lr": 1.3236587293339974e-05, "epoch": 1.9445506692160612, "percentage": 64.94, "elapsed_time": "1:31:45", "remaining_time": "0:49:32"}
+{"current_steps": 340, "total_steps": 522, "loss": 0.4238, "lr": 1.3110662994313343e-05, "epoch": 1.9502868068833652, "percentage": 65.13, "elapsed_time": "1:32:01", "remaining_time": "0:49:15"}
+{"current_steps": 341, "total_steps": 522, "loss": 0.4284, "lr": 1.2985047817251916e-05, "epoch": 1.9560229445506692, "percentage": 65.33, "elapsed_time": "1:32:18", "remaining_time": "0:48:59"}
+{"current_steps": 342, "total_steps": 522, "loss": 0.4178, "lr": 1.2859747398461542e-05, "epoch": 1.9617590822179731, "percentage": 65.52, "elapsed_time": "1:32:33", "remaining_time": "0:48:42"}
+{"current_steps": 343, "total_steps": 522, "loss": 0.3667, "lr": 1.2734767360124955e-05, "epoch": 1.967495219885277, "percentage": 65.71, "elapsed_time": "1:32:43", "remaining_time": "0:48:23"}
+{"current_steps": 344, "total_steps": 522, "loss": 0.3512, "lr": 1.261011331004958e-05, "epoch": 1.9732313575525813, "percentage": 65.9, "elapsed_time": "1:32:56", "remaining_time": "0:48:05"}
+{"current_steps": 345, "total_steps": 522, "loss": 0.4461, "lr": 1.2485790841415834e-05, "epoch": 1.9789674952198852, "percentage": 66.09, "elapsed_time": "1:33:12", "remaining_time": "0:47:49"}
+{"current_steps": 346, "total_steps": 522, "loss": 0.4303, "lr": 1.2361805532526225e-05, "epoch": 1.9847036328871894, "percentage": 66.28, "elapsed_time": "1:33:30", "remaining_time": "0:47:34"}
+{"current_steps": 347, "total_steps": 522, "loss": 0.4182, "lr": 1.2238162946555004e-05, "epoch": 1.9904397705544934, "percentage": 66.48, "elapsed_time": "1:33:45", "remaining_time": "0:47:17"}
+{"current_steps": 348, "total_steps": 522, "loss": 0.3843, "lr": 1.2114868631298608e-05, "epoch": 1.9961759082217974, "percentage": 66.67, "elapsed_time": "1:34:03", "remaining_time": "0:47:01"}
+{"current_steps": 349, "total_steps": 522, "loss": 0.5702, "lr": 1.1991928118926661e-05, "epoch": 2.0019120458891013, "percentage": 66.86, "elapsed_time": "1:35:07", "remaining_time": "0:47:09"}
+{"current_steps": 350, "total_steps": 522, "loss": 0.3478, "lr": 1.1869346925733813e-05, "epoch": 2.0076481835564053, "percentage": 67.05, "elapsed_time": "1:35:23", "remaining_time": "0:46:52"}
+{"current_steps": 351, "total_steps": 522, "loss": 0.3504, "lr": 1.1747130551892176e-05, "epoch": 2.0133843212237093, "percentage": 67.24, "elapsed_time": "1:35:39", "remaining_time": "0:46:35"}
+{"current_steps": 352, "total_steps": 522, "loss": 0.3247, "lr": 1.1625284481204577e-05, "epoch": 2.019120458891013, "percentage": 67.43, "elapsed_time": "1:35:52", "remaining_time": "0:46:18"}
+{"current_steps": 353, "total_steps": 522, "loss": 0.3293, "lr": 1.1503814180858451e-05, "epoch": 2.024856596558317, "percentage": 67.62, "elapsed_time": "1:36:09", "remaining_time": "0:46:01"}
+{"current_steps": 354, "total_steps": 522, "loss": 0.3068, "lr": 1.1382725101180593e-05, "epoch": 2.0305927342256216, "percentage": 67.82, "elapsed_time": "1:36:25", "remaining_time": "0:45:45"}
+{"current_steps": 355, "total_steps": 522, "loss": 0.3031, "lr": 1.1262022675392544e-05, "epoch": 2.0363288718929256, "percentage": 68.01, "elapsed_time": "1:36:44", "remaining_time": "0:45:30"}
+{"current_steps": 356, "total_steps": 522, "loss": 0.3104, "lr": 1.1141712319366835e-05, "epoch": 2.0420650095602295, "percentage": 68.2, "elapsed_time": "1:36:59", "remaining_time": "0:45:13"}
+{"current_steps": 357, "total_steps": 522, "loss": 0.3355, "lr": 1.1021799431383969e-05, "epoch": 2.0478011472275335, "percentage": 68.39, "elapsed_time": "1:37:18", "remaining_time": "0:44:58"}
+{"current_steps": 358, "total_steps": 522, "loss": 0.3181, "lr": 1.0902289391890232e-05, "epoch": 2.0535372848948374, "percentage": 68.58, "elapsed_time": "1:37:38", "remaining_time": "0:44:43"}