Training in progress, epoch 3

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:44aae7c0bf68809018d29ed21ee9663f653e1b06e87096e237b77f3e182ed829
+oid sha256:26964b67ed54f423824d2df494fa5d9689fee883d17730b4f17d6e039a2ba415
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9f30432b00d5df5aec6d69c455961d81992ed9666e17bfe3ebf842a1ee544c2d
+oid sha256:ff82cb342cababce411bcc926bd14d6d12bfa5faac740ec30c766ab1d732f99f
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:51b7e1042423d5ba4ee9583caba4c819ee382fcede77f85ca6913477fe37c288
+oid sha256:61005b8c7ae26045005af67965dfa2923768f8d07e2421d6b99609dff5c3ad41
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d407b5eaf3c79e959a512915058eee058d0e673e7ed60ccc0e4f3671868e148b
+oid sha256:271eae1a99201b1f99c07ca295ac1b323a68795c69f2208ec3b226c39f211c92
 size 1089994880

trainer_log.jsonl

@@ -538,3 +538,183 @@
 {"current_steps": 538, "total_steps": 895, "loss": 0.7202, "lr": 3.2934535420456506e-05, "epoch": 2.9909659485753997, "percentage": 60.11, "elapsed_time": "15:23:07", "remaining_time": "10:12:33"}
 {"current_steps": 539, "total_steps": 895, "loss": 0.7312, "lr": 3.2780940190365043e-05, "epoch": 2.996525364836692, "percentage": 60.22, "elapsed_time": "15:24:49", "remaining_time": "10:10:49"}
 {"current_steps": 540, "total_steps": 895, "loss": 0.725, "lr": 3.262745490846084e-05, "epoch": 3.0020847810979845, "percentage": 60.34, "elapsed_time": "15:28:20", "remaining_time": "10:10:18"}
+{"current_steps": 541, "total_steps": 895, "loss": 0.6976, "lr": 3.2474081912365245e-05, "epoch": 3.0076441973592773, "percentage": 60.45, "elapsed_time": "15:30:03", "remaining_time": "10:08:34"}
+{"current_steps": 542, "total_steps": 895, "loss": 0.7119, "lr": 3.23208235379895e-05, "epoch": 3.0132036136205698, "percentage": 60.56, "elapsed_time": "15:31:45", "remaining_time": "10:06:50"}
+{"current_steps": 543, "total_steps": 895, "loss": 0.7182, "lr": 3.216768211949912e-05, "epoch": 3.018763029881862, "percentage": 60.67, "elapsed_time": "15:33:28", "remaining_time": "10:05:07"}
+{"current_steps": 544, "total_steps": 895, "loss": 0.7012, "lr": 3.201465998927835e-05, "epoch": 3.024322446143155, "percentage": 60.78, "elapsed_time": "15:35:10", "remaining_time": "10:03:23"}
+{"current_steps": 545, "total_steps": 895, "loss": 0.6971, "lr": 3.1861759477894656e-05, "epoch": 3.0298818624044475, "percentage": 60.89, "elapsed_time": "15:36:51", "remaining_time": "10:01:39"}
+{"current_steps": 546, "total_steps": 895, "loss": 0.7005, "lr": 3.170898291406317e-05, "epoch": 3.03544127866574, "percentage": 61.01, "elapsed_time": "15:38:34", "remaining_time": "9:59:55"}
+{"current_steps": 547, "total_steps": 895, "loss": 0.692, "lr": 3.155633262461133e-05, "epoch": 3.0410006949270327, "percentage": 61.12, "elapsed_time": "15:40:16", "remaining_time": "9:58:11"}
+{"current_steps": 548, "total_steps": 895, "loss": 0.6979, "lr": 3.1403810934443346e-05, "epoch": 3.046560111188325, "percentage": 61.23, "elapsed_time": "15:41:56", "remaining_time": "9:56:27"}
+{"current_steps": 549, "total_steps": 895, "loss": 0.6943, "lr": 3.125142016650482e-05, "epoch": 3.052119527449618, "percentage": 61.34, "elapsed_time": "15:43:37", "remaining_time": "9:54:42"}
+{"current_steps": 550, "total_steps": 895, "loss": 0.6992, "lr": 3.109916264174743e-05, "epoch": 3.0576789437109104, "percentage": 61.45, "elapsed_time": "15:45:20", "remaining_time": "9:52:59"}
+{"current_steps": 551, "total_steps": 895, "loss": 0.7074, "lr": 3.094704067909345e-05, "epoch": 3.063238359972203, "percentage": 61.56, "elapsed_time": "15:47:02", "remaining_time": "9:51:15"}
+{"current_steps": 552, "total_steps": 895, "loss": 0.695, "lr": 3.079505659540055e-05, "epoch": 3.0687977762334953, "percentage": 61.68, "elapsed_time": "15:48:44", "remaining_time": "9:49:31"}
+{"current_steps": 553, "total_steps": 895, "loss": 0.7091, "lr": 3.064321270542646e-05, "epoch": 3.074357192494788, "percentage": 61.79, "elapsed_time": "15:50:26", "remaining_time": "9:47:47"}
+{"current_steps": 554, "total_steps": 895, "loss": 0.7176, "lr": 3.0491511321793737e-05, "epoch": 3.0799166087560805, "percentage": 61.9, "elapsed_time": "15:52:08", "remaining_time": "9:46:03"}
+{"current_steps": 555, "total_steps": 895, "loss": 0.7092, "lr": 3.0339954754954488e-05, "epoch": 3.085476025017373, "percentage": 62.01, "elapsed_time": "15:53:49", "remaining_time": "9:44:19"}
+{"current_steps": 556, "total_steps": 895, "loss": 0.7044, "lr": 3.0188545313155257e-05, "epoch": 3.091035441278666, "percentage": 62.12, "elapsed_time": "15:55:31", "remaining_time": "9:42:35"}
+{"current_steps": 557, "total_steps": 895, "loss": 0.7107, "lr": 3.0037285302401828e-05, "epoch": 3.0965948575399582, "percentage": 62.23, "elapsed_time": "15:57:14", "remaining_time": "9:40:52"}
+{"current_steps": 558, "total_steps": 895, "loss": 0.7063, "lr": 2.9886177026424107e-05, "epoch": 3.102154273801251, "percentage": 62.35, "elapsed_time": "15:58:57", "remaining_time": "9:39:09"}
+{"current_steps": 559, "total_steps": 895, "loss": 0.6995, "lr": 2.9735222786641067e-05, "epoch": 3.1077136900625435, "percentage": 62.46, "elapsed_time": "16:00:39", "remaining_time": "9:37:25"}
+{"current_steps": 560, "total_steps": 895, "loss": 0.7149, "lr": 2.9584424882125593e-05, "epoch": 3.113273106323836, "percentage": 62.57, "elapsed_time": "16:02:21", "remaining_time": "9:35:41"}
+{"current_steps": 561, "total_steps": 895, "loss": 0.7155, "lr": 2.943378560956962e-05, "epoch": 3.1188325225851288, "percentage": 62.68, "elapsed_time": "16:04:03", "remaining_time": "9:33:57"}
+{"current_steps": 562, "total_steps": 895, "loss": 0.7024, "lr": 2.9283307263249048e-05, "epoch": 3.124391938846421, "percentage": 62.79, "elapsed_time": "16:05:45", "remaining_time": "9:32:13"}
+{"current_steps": 563, "total_steps": 895, "loss": 0.6945, "lr": 2.913299213498884e-05, "epoch": 3.1299513551077136, "percentage": 62.91, "elapsed_time": "16:07:27", "remaining_time": "9:30:30"}
+{"current_steps": 564, "total_steps": 895, "loss": 0.7007, "lr": 2.8982842514128058e-05, "epoch": 3.135510771369006, "percentage": 63.02, "elapsed_time": "16:09:08", "remaining_time": "9:28:46"}
+{"current_steps": 565, "total_steps": 895, "loss": 0.7113, "lr": 2.8832860687485107e-05, "epoch": 3.141070187630299, "percentage": 63.13, "elapsed_time": "16:10:49", "remaining_time": "9:27:01"}
+{"current_steps": 566, "total_steps": 895, "loss": 0.679, "lr": 2.86830489393228e-05, "epoch": 3.1466296038915913, "percentage": 63.24, "elapsed_time": "16:12:32", "remaining_time": "9:25:18"}
+{"current_steps": 567, "total_steps": 895, "loss": 0.7127, "lr": 2.8533409551313628e-05, "epoch": 3.1521890201528837, "percentage": 63.35, "elapsed_time": "16:14:14", "remaining_time": "9:23:34"}
+{"current_steps": 568, "total_steps": 895, "loss": 0.6942, "lr": 2.8383944802505003e-05, "epoch": 3.1577484364141766, "percentage": 63.46, "elapsed_time": "16:15:58", "remaining_time": "9:21:52"}
+{"current_steps": 569, "total_steps": 895, "loss": 0.6891, "lr": 2.8234656969284483e-05, "epoch": 3.163307852675469, "percentage": 63.58, "elapsed_time": "16:17:40", "remaining_time": "9:20:08"}
+{"current_steps": 570, "total_steps": 895, "loss": 0.6943, "lr": 2.8085548325345208e-05, "epoch": 3.1688672689367614, "percentage": 63.69, "elapsed_time": "16:19:22", "remaining_time": "9:18:24"}
+{"current_steps": 571, "total_steps": 895, "loss": 0.7037, "lr": 2.7936621141651197e-05, "epoch": 3.1744266851980543, "percentage": 63.8, "elapsed_time": "16:21:04", "remaining_time": "9:16:41"}
+{"current_steps": 572, "total_steps": 895, "loss": 0.6951, "lr": 2.7787877686402807e-05, "epoch": 3.1799861014593467, "percentage": 63.91, "elapsed_time": "16:22:45", "remaining_time": "9:14:56"}
+{"current_steps": 573, "total_steps": 895, "loss": 0.6918, "lr": 2.7639320225002108e-05, "epoch": 3.185545517720639, "percentage": 64.02, "elapsed_time": "16:24:26", "remaining_time": "9:13:12"}
+{"current_steps": 574, "total_steps": 895, "loss": 0.6989, "lr": 2.7490951020018476e-05, "epoch": 3.191104933981932, "percentage": 64.13, "elapsed_time": "16:26:08", "remaining_time": "9:11:28"}
+{"current_steps": 575, "total_steps": 895, "loss": 0.6982, "lr": 2.7342772331154102e-05, "epoch": 3.1966643502432244, "percentage": 64.25, "elapsed_time": "16:27:51", "remaining_time": "9:09:45"}
+{"current_steps": 576, "total_steps": 895, "loss": 0.6984, "lr": 2.7194786415209528e-05, "epoch": 3.2022237665045172, "percentage": 64.36, "elapsed_time": "16:29:34", "remaining_time": "9:08:02"}
+{"current_steps": 577, "total_steps": 895, "loss": 0.7081, "lr": 2.7046995526049334e-05, "epoch": 3.2077831827658096, "percentage": 64.47, "elapsed_time": "16:31:17", "remaining_time": "9:06:19"}
+{"current_steps": 578, "total_steps": 895, "loss": 0.7005, "lr": 2.68994019145678e-05, "epoch": 3.213342599027102, "percentage": 64.58, "elapsed_time": "16:32:58", "remaining_time": "9:04:35"}
+{"current_steps": 579, "total_steps": 895, "loss": 0.7108, "lr": 2.675200782865459e-05, "epoch": 3.218902015288395, "percentage": 64.69, "elapsed_time": "16:34:40", "remaining_time": "9:02:51"}
+{"current_steps": 580, "total_steps": 895, "loss": 0.704, "lr": 2.6604815513160556e-05, "epoch": 3.2244614315496873, "percentage": 64.8, "elapsed_time": "16:36:22", "remaining_time": "9:01:07"}
+{"current_steps": 581, "total_steps": 895, "loss": 0.712, "lr": 2.645782720986356e-05, "epoch": 3.2300208478109798, "percentage": 64.92, "elapsed_time": "16:38:03", "remaining_time": "8:59:24"}
+{"current_steps": 582, "total_steps": 895, "loss": 0.689, "lr": 2.6311045157434235e-05, "epoch": 3.2355802640722726, "percentage": 65.03, "elapsed_time": "16:39:44", "remaining_time": "8:57:39"}
+{"current_steps": 583, "total_steps": 895, "loss": 0.716, "lr": 2.6164471591402016e-05, "epoch": 3.241139680333565, "percentage": 65.14, "elapsed_time": "16:41:26", "remaining_time": "8:55:56"}
+{"current_steps": 584, "total_steps": 895, "loss": 0.7085, "lr": 2.6018108744121032e-05, "epoch": 3.2466990965948574, "percentage": 65.25, "elapsed_time": "16:43:11", "remaining_time": "8:54:13"}
+{"current_steps": 585, "total_steps": 895, "loss": 0.7079, "lr": 2.5871958844736083e-05, "epoch": 3.25225851285615, "percentage": 65.36, "elapsed_time": "16:44:55", "remaining_time": "8:52:31"}
+{"current_steps": 586, "total_steps": 895, "loss": 0.7156, "lr": 2.5726024119148757e-05, "epoch": 3.2578179291174427, "percentage": 65.47, "elapsed_time": "16:46:38", "remaining_time": "8:50:48"}
+{"current_steps": 587, "total_steps": 895, "loss": 0.7082, "lr": 2.5580306789983418e-05, "epoch": 3.263377345378735, "percentage": 65.59, "elapsed_time": "16:48:21", "remaining_time": "8:49:05"}
+{"current_steps": 588, "total_steps": 895, "loss": 0.6969, "lr": 2.54348090765535e-05, "epoch": 3.2689367616400276, "percentage": 65.7, "elapsed_time": "16:50:04", "remaining_time": "8:47:22"}
+{"current_steps": 589, "total_steps": 895, "loss": 0.6864, "lr": 2.5289533194827586e-05, "epoch": 3.2744961779013204, "percentage": 65.81, "elapsed_time": "16:51:48", "remaining_time": "8:45:39"}
+{"current_steps": 590, "total_steps": 895, "loss": 0.7149, "lr": 2.514448135739574e-05, "epoch": 3.280055594162613, "percentage": 65.92, "elapsed_time": "16:53:32", "remaining_time": "8:43:56"}
+{"current_steps": 591, "total_steps": 895, "loss": 0.6965, "lr": 2.4999655773435726e-05, "epoch": 3.2856150104239052, "percentage": 66.03, "elapsed_time": "16:55:14", "remaining_time": "8:42:13"}
+{"current_steps": 592, "total_steps": 895, "loss": 0.709, "lr": 2.4855058648679434e-05, "epoch": 3.291174426685198, "percentage": 66.15, "elapsed_time": "16:56:57", "remaining_time": "8:40:30"}
+{"current_steps": 593, "total_steps": 895, "loss": 0.6981, "lr": 2.4710692185379264e-05, "epoch": 3.2967338429464905, "percentage": 66.26, "elapsed_time": "16:58:40", "remaining_time": "8:38:47"}
+{"current_steps": 594, "total_steps": 895, "loss": 0.7, "lr": 2.4566558582274583e-05, "epoch": 3.302293259207783, "percentage": 66.37, "elapsed_time": "17:00:22", "remaining_time": "8:37:03"}
+{"current_steps": 595, "total_steps": 895, "loss": 0.6986, "lr": 2.442266003455823e-05, "epoch": 3.307852675469076, "percentage": 66.48, "elapsed_time": "17:02:05", "remaining_time": "8:35:20"}
+{"current_steps": 596, "total_steps": 895, "loss": 0.7138, "lr": 2.427899873384306e-05, "epoch": 3.313412091730368, "percentage": 66.59, "elapsed_time": "17:03:48", "remaining_time": "8:33:37"}
+{"current_steps": 597, "total_steps": 895, "loss": 0.7013, "lr": 2.4135576868128636e-05, "epoch": 3.3189715079916606, "percentage": 66.7, "elapsed_time": "17:05:29", "remaining_time": "8:31:53"}
+{"current_steps": 598, "total_steps": 895, "loss": 0.7144, "lr": 2.399239662176785e-05, "epoch": 3.3245309242529535, "percentage": 66.82, "elapsed_time": "17:07:11", "remaining_time": "8:30:09"}
+{"current_steps": 599, "total_steps": 895, "loss": 0.716, "lr": 2.3849460175433673e-05, "epoch": 3.330090340514246, "percentage": 66.93, "elapsed_time": "17:08:53", "remaining_time": "8:28:26"}
+{"current_steps": 600, "total_steps": 895, "loss": 0.6947, "lr": 2.3706769706085896e-05, "epoch": 3.3356497567755383, "percentage": 67.04, "elapsed_time": "17:10:36", "remaining_time": "8:26:42"}
+{"current_steps": 601, "total_steps": 895, "loss": 0.7044, "lr": 2.3564327386938075e-05, "epoch": 3.341209173036831, "percentage": 67.15, "elapsed_time": "17:12:20", "remaining_time": "8:25:00"}
+{"current_steps": 602, "total_steps": 895, "loss": 0.6948, "lr": 2.3422135387424342e-05, "epoch": 3.3467685892981236, "percentage": 67.26, "elapsed_time": "17:14:03", "remaining_time": "8:23:17"}
+{"current_steps": 603, "total_steps": 895, "loss": 0.7178, "lr": 2.3280195873166384e-05, "epoch": 3.352328005559416, "percentage": 67.37, "elapsed_time": "17:15:47", "remaining_time": "8:21:34"}
+{"current_steps": 604, "total_steps": 895, "loss": 0.6903, "lr": 2.3138511005940502e-05, "epoch": 3.357887421820709, "percentage": 67.49, "elapsed_time": "17:17:30", "remaining_time": "8:19:51"}
+{"current_steps": 605, "total_steps": 895, "loss": 0.6961, "lr": 2.2997082943644635e-05, "epoch": 3.3634468380820013, "percentage": 67.6, "elapsed_time": "17:19:14", "remaining_time": "8:18:08"}
+{"current_steps": 606, "total_steps": 895, "loss": 0.716, "lr": 2.2855913840265504e-05, "epoch": 3.3690062543432937, "percentage": 67.71, "elapsed_time": "17:20:56", "remaining_time": "8:16:25"}
+{"current_steps": 607, "total_steps": 895, "loss": 0.6921, "lr": 2.271500584584585e-05, "epoch": 3.3745656706045866, "percentage": 67.82, "elapsed_time": "17:22:38", "remaining_time": "8:14:41"}
+{"current_steps": 608, "total_steps": 895, "loss": 0.6911, "lr": 2.2574361106451637e-05, "epoch": 3.380125086865879, "percentage": 67.93, "elapsed_time": "17:24:20", "remaining_time": "8:12:58"}
+{"current_steps": 609, "total_steps": 895, "loss": 0.7096, "lr": 2.2433981764139383e-05, "epoch": 3.385684503127172, "percentage": 68.04, "elapsed_time": "17:26:03", "remaining_time": "8:11:14"}
+{"current_steps": 610, "total_steps": 895, "loss": 0.6982, "lr": 2.2293869956923517e-05, "epoch": 3.3912439193884643, "percentage": 68.16, "elapsed_time": "17:27:44", "remaining_time": "8:09:31"}
+{"current_steps": 611, "total_steps": 895, "loss": 0.701, "lr": 2.215402781874391e-05, "epoch": 3.3968033356497567, "percentage": 68.27, "elapsed_time": "17:29:25", "remaining_time": "8:07:47"}
+{"current_steps": 612, "total_steps": 895, "loss": 0.7006, "lr": 2.2014457479433204e-05, "epoch": 3.4023627519110495, "percentage": 68.38, "elapsed_time": "17:31:06", "remaining_time": "8:06:03"}
+{"current_steps": 613, "total_steps": 895, "loss": 0.7011, "lr": 2.1875161064684586e-05, "epoch": 3.407922168172342, "percentage": 68.49, "elapsed_time": "17:32:48", "remaining_time": "8:04:19"}
+{"current_steps": 614, "total_steps": 895, "loss": 0.6973, "lr": 2.1736140696019166e-05, "epoch": 3.4134815844336344, "percentage": 68.6, "elapsed_time": "17:34:31", "remaining_time": "8:02:36"}
+{"current_steps": 615, "total_steps": 895, "loss": 0.7066, "lr": 2.1597398490753917e-05, "epoch": 3.4190410006949272, "percentage": 68.72, "elapsed_time": "17:36:13", "remaining_time": "8:00:52"}
+{"current_steps": 616, "total_steps": 895, "loss": 0.7026, "lr": 2.1458936561969254e-05, "epoch": 3.4246004169562196, "percentage": 68.83, "elapsed_time": "17:37:55", "remaining_time": "7:59:09"}
+{"current_steps": 617, "total_steps": 895, "loss": 0.708, "lr": 2.132075701847691e-05, "epoch": 3.430159833217512, "percentage": 68.94, "elapsed_time": "17:39:37", "remaining_time": "7:57:26"}
+{"current_steps": 618, "total_steps": 895, "loss": 0.7115, "lr": 2.1182861964787803e-05, "epoch": 3.435719249478805, "percentage": 69.05, "elapsed_time": "17:41:19", "remaining_time": "7:55:42"}
+{"current_steps": 619, "total_steps": 895, "loss": 0.6971, "lr": 2.1045253501080058e-05, "epoch": 3.4412786657400973, "percentage": 69.16, "elapsed_time": "17:43:00", "remaining_time": "7:53:58"}
+{"current_steps": 620, "total_steps": 895, "loss": 0.7074, "lr": 2.0907933723166888e-05, "epoch": 3.4468380820013897, "percentage": 69.27, "elapsed_time": "17:44:42", "remaining_time": "7:52:14"}
+{"current_steps": 621, "total_steps": 895, "loss": 0.6917, "lr": 2.0770904722464764e-05, "epoch": 3.4523974982626826, "percentage": 69.39, "elapsed_time": "17:46:25", "remaining_time": "7:50:31"}
+{"current_steps": 622, "total_steps": 895, "loss": 0.7093, "lr": 2.06341685859616e-05, "epoch": 3.457956914523975, "percentage": 69.5, "elapsed_time": "17:48:07", "remaining_time": "7:48:48"}
+{"current_steps": 623, "total_steps": 895, "loss": 0.7032, "lr": 2.049772739618479e-05, "epoch": 3.4635163307852674, "percentage": 69.61, "elapsed_time": "17:49:51", "remaining_time": "7:47:05"}
+{"current_steps": 624, "total_steps": 895, "loss": 0.709, "lr": 2.0361583231169747e-05, "epoch": 3.4690757470465603, "percentage": 69.72, "elapsed_time": "17:51:33", "remaining_time": "7:45:22"}
+{"current_steps": 625, "total_steps": 895, "loss": 0.6936, "lr": 2.022573816442801e-05, "epoch": 3.4746351633078527, "percentage": 69.83, "elapsed_time": "17:53:14", "remaining_time": "7:43:38"}
+{"current_steps": 626, "total_steps": 895, "loss": 0.7118, "lr": 2.0090194264915873e-05, "epoch": 3.480194579569145, "percentage": 69.94, "elapsed_time": "17:54:56", "remaining_time": "7:41:54"}
+{"current_steps": 627, "total_steps": 895, "loss": 0.7105, "lr": 1.9954953597002654e-05, "epoch": 3.485753995830438, "percentage": 70.06, "elapsed_time": "17:56:38", "remaining_time": "7:40:11"}
+{"current_steps": 628, "total_steps": 895, "loss": 0.6957, "lr": 1.9820018220439487e-05, "epoch": 3.4913134120917304, "percentage": 70.17, "elapsed_time": "17:58:20", "remaining_time": "7:38:27"}
+{"current_steps": 629, "total_steps": 895, "loss": 0.7098, "lr": 1.9685390190327757e-05, "epoch": 3.496872828353023, "percentage": 70.28, "elapsed_time": "18:00:01", "remaining_time": "7:36:44"}
+{"current_steps": 630, "total_steps": 895, "loss": 0.6971, "lr": 1.955107155708795e-05, "epoch": 3.5024322446143152, "percentage": 70.39, "elapsed_time": "18:01:44", "remaining_time": "7:35:01"}
+{"current_steps": 631, "total_steps": 895, "loss": 0.7186, "lr": 1.9417064366428295e-05, "epoch": 3.507991660875608, "percentage": 70.5, "elapsed_time": "18:03:27", "remaining_time": "7:33:18"}
+{"current_steps": 632, "total_steps": 895, "loss": 0.6964, "lr": 1.92833706593137e-05, "epoch": 3.5135510771369005, "percentage": 70.61, "elapsed_time": "18:05:11", "remaining_time": "7:31:35"}
+{"current_steps": 633, "total_steps": 895, "loss": 0.6896, "lr": 1.9149992471934614e-05, "epoch": 3.519110493398193, "percentage": 70.73, "elapsed_time": "18:06:54", "remaining_time": "7:29:52"}
+{"current_steps": 634, "total_steps": 895, "loss": 0.705, "lr": 1.9016931835676078e-05, "epoch": 3.524669909659486, "percentage": 70.84, "elapsed_time": "18:08:36", "remaining_time": "7:28:08"}
+{"current_steps": 635, "total_steps": 895, "loss": 0.6974, "lr": 1.8884190777086705e-05, "epoch": 3.530229325920778, "percentage": 70.95, "elapsed_time": "18:10:19", "remaining_time": "7:26:25"}
+{"current_steps": 636, "total_steps": 895, "loss": 0.706, "lr": 1.8751771317847876e-05, "epoch": 3.5357887421820706, "percentage": 71.06, "elapsed_time": "18:12:01", "remaining_time": "7:24:42"}
+{"current_steps": 637, "total_steps": 895, "loss": 0.7115, "lr": 1.8619675474742904e-05, "epoch": 3.5413481584433635, "percentage": 71.17, "elapsed_time": "18:13:42", "remaining_time": "7:22:58"}
+{"current_steps": 638, "total_steps": 895, "loss": 0.7107, "lr": 1.848790525962639e-05, "epoch": 3.546907574704656, "percentage": 71.28, "elapsed_time": "18:15:23", "remaining_time": "7:21:14"}
+{"current_steps": 639, "total_steps": 895, "loss": 0.7102, "lr": 1.8356462679393488e-05, "epoch": 3.5524669909659483, "percentage": 71.4, "elapsed_time": "18:17:06", "remaining_time": "7:19:31"}
+{"current_steps": 640, "total_steps": 895, "loss": 0.7051, "lr": 1.8225349735949435e-05, "epoch": 3.558026407227241, "percentage": 71.51, "elapsed_time": "18:18:48", "remaining_time": "7:17:48"}
+{"current_steps": 641, "total_steps": 895, "loss": 0.7093, "lr": 1.809456842617893e-05, "epoch": 3.5635858234885336, "percentage": 71.62, "elapsed_time": "18:20:30", "remaining_time": "7:16:05"}
+{"current_steps": 642, "total_steps": 895, "loss": 0.7072, "lr": 1.7964120741915905e-05, "epoch": 3.569145239749826, "percentage": 71.73, "elapsed_time": "18:22:13", "remaining_time": "7:14:21"}
+{"current_steps": 643, "total_steps": 895, "loss": 0.7013, "lr": 1.7834008669913016e-05, "epoch": 3.574704656011119, "percentage": 71.84, "elapsed_time": "18:23:56", "remaining_time": "7:12:39"}
+{"current_steps": 644, "total_steps": 895, "loss": 0.711, "lr": 1.770423419181147e-05, "epoch": 3.5802640722724113, "percentage": 71.96, "elapsed_time": "18:25:39", "remaining_time": "7:10:55"}
+{"current_steps": 645, "total_steps": 895, "loss": 0.6973, "lr": 1.7574799284110863e-05, "epoch": 3.5858234885337037, "percentage": 72.07, "elapsed_time": "18:27:20", "remaining_time": "7:09:12"}
+{"current_steps": 646, "total_steps": 895, "loss": 0.6946, "lr": 1.744570591813901e-05, "epoch": 3.5913829047949966, "percentage": 72.18, "elapsed_time": "18:29:01", "remaining_time": "7:07:28"}
+{"current_steps": 647, "total_steps": 895, "loss": 0.6827, "lr": 1.7316956060021957e-05, "epoch": 3.596942321056289, "percentage": 72.29, "elapsed_time": "18:30:42", "remaining_time": "7:05:44"}
+{"current_steps": 648, "total_steps": 895, "loss": 0.7036, "lr": 1.718855167065404e-05, "epoch": 3.6025017373175814, "percentage": 72.4, "elapsed_time": "18:32:24", "remaining_time": "7:04:01"}
+{"current_steps": 649, "total_steps": 895, "loss": 0.6964, "lr": 1.706049470566803e-05, "epoch": 3.6080611535788742, "percentage": 72.51, "elapsed_time": "18:34:06", "remaining_time": "7:02:17"}
+{"current_steps": 650, "total_steps": 895, "loss": 0.703, "lr": 1.6932787115405318e-05, "epoch": 3.6136205698401667, "percentage": 72.63, "elapsed_time": "18:35:50", "remaining_time": "7:00:35"}
+{"current_steps": 651, "total_steps": 895, "loss": 0.7104, "lr": 1.6805430844886226e-05, "epoch": 3.619179986101459, "percentage": 72.74, "elapsed_time": "18:37:32", "remaining_time": "6:58:51"}
+{"current_steps": 652, "total_steps": 895, "loss": 0.7014, "lr": 1.667842783378036e-05, "epoch": 3.624739402362752, "percentage": 72.85, "elapsed_time": "18:39:15", "remaining_time": "6:57:08"}
+{"current_steps": 653, "total_steps": 895, "loss": 0.697, "lr": 1.655178001637716e-05, "epoch": 3.6302988186240444, "percentage": 72.96, "elapsed_time": "18:40:58", "remaining_time": "6:55:25"}
+{"current_steps": 654, "total_steps": 895, "loss": 0.7171, "lr": 1.6425489321556306e-05, "epoch": 3.6358582348853368, "percentage": 73.07, "elapsed_time": "18:42:40", "remaining_time": "6:53:42"}
+{"current_steps": 655, "total_steps": 895, "loss": 0.7063, "lr": 1.6299557672758434e-05, "epoch": 3.6414176511466296, "percentage": 73.18, "elapsed_time": "18:44:22", "remaining_time": "6:51:58"}
+{"current_steps": 656, "total_steps": 895, "loss": 0.7014, "lr": 1.6173986987955794e-05, "epoch": 3.646977067407922, "percentage": 73.3, "elapsed_time": "18:46:04", "remaining_time": "6:50:15"}
+{"current_steps": 657, "total_steps": 895, "loss": 0.6987, "lr": 1.6048779179623098e-05, "epoch": 3.6525364836692145, "percentage": 73.41, "elapsed_time": "18:47:46", "remaining_time": "6:48:32"}
+{"current_steps": 658, "total_steps": 895, "loss": 0.6992, "lr": 1.5923936154708332e-05, "epoch": 3.6580958999305073, "percentage": 73.52, "elapsed_time": "18:49:27", "remaining_time": "6:46:48"}
+{"current_steps": 659, "total_steps": 895, "loss": 0.6939, "lr": 1.5799459814603716e-05, "epoch": 3.6636553161917997, "percentage": 73.63, "elapsed_time": "18:51:09", "remaining_time": "6:45:05"}
+{"current_steps": 660, "total_steps": 895, "loss": 0.6972, "lr": 1.5675352055116774e-05, "epoch": 3.669214732453092, "percentage": 73.74, "elapsed_time": "18:52:52", "remaining_time": "6:43:22"}
+{"current_steps": 661, "total_steps": 895, "loss": 0.691, "lr": 1.5551614766441483e-05, "epoch": 3.674774148714385, "percentage": 73.85, "elapsed_time": "18:54:36", "remaining_time": "6:41:39"}
+{"current_steps": 662, "total_steps": 895, "loss": 0.6931, "lr": 1.54282498331294e-05, "epoch": 3.6803335649756774, "percentage": 73.97, "elapsed_time": "18:56:19", "remaining_time": "6:39:56"}
+{"current_steps": 663, "total_steps": 895, "loss": 0.6965, "lr": 1.5305259134061075e-05, "epoch": 3.68589298123697, "percentage": 74.08, "elapsed_time": "18:58:03", "remaining_time": "6:38:14"}
+{"current_steps": 664, "total_steps": 895, "loss": 0.7041, "lr": 1.5182644542417278e-05, "epoch": 3.6914523974982627, "percentage": 74.19, "elapsed_time": "18:59:45", "remaining_time": "6:36:30"}
+{"current_steps": 665, "total_steps": 895, "loss": 0.6907, "lr": 1.5060407925650662e-05, "epoch": 3.697011813759555, "percentage": 74.3, "elapsed_time": "19:01:26", "remaining_time": "6:34:46"}
+{"current_steps": 666, "total_steps": 895, "loss": 0.6991, "lr": 1.4938551145457152e-05, "epoch": 3.7025712300208475, "percentage": 74.41, "elapsed_time": "19:03:08", "remaining_time": "6:33:03"}
+{"current_steps": 667, "total_steps": 895, "loss": 0.7089, "lr": 1.4817076057747728e-05, "epoch": 3.7081306462821404, "percentage": 74.53, "elapsed_time": "19:04:50", "remaining_time": "6:31:20"}
+{"current_steps": 668, "total_steps": 895, "loss": 0.7191, "lr": 1.4695984512619999e-05, "epoch": 3.713690062543433, "percentage": 74.64, "elapsed_time": "19:06:31", "remaining_time": "6:29:36"}
+{"current_steps": 669, "total_steps": 895, "loss": 0.6991, "lr": 1.4575278354330213e-05, "epoch": 3.7192494788047252, "percentage": 74.75, "elapsed_time": "19:08:12", "remaining_time": "6:27:53"}
+{"current_steps": 670, "total_steps": 895, "loss": 0.7049, "lr": 1.4454959421265029e-05, "epoch": 3.724808895066018, "percentage": 74.86, "elapsed_time": "19:09:57", "remaining_time": "6:26:10"}
+{"current_steps": 671, "total_steps": 895, "loss": 0.7065, "lr": 1.4335029545913556e-05, "epoch": 3.7303683113273105, "percentage": 74.97, "elapsed_time": "19:11:39", "remaining_time": "6:24:27"}
+{"current_steps": 672, "total_steps": 895, "loss": 0.6959, "lr": 1.4215490554839483e-05, "epoch": 3.735927727588603, "percentage": 75.08, "elapsed_time": "19:13:22", "remaining_time": "6:22:44"}
+{"current_steps": 673, "total_steps": 895, "loss": 0.6953, "lr": 1.409634426865321e-05, "epoch": 3.741487143849896, "percentage": 75.2, "elapsed_time": "19:15:06", "remaining_time": "6:21:01"}
+{"current_steps": 674, "total_steps": 895, "loss": 0.7022, "lr": 1.397759250198413e-05, "epoch": 3.747046560111188, "percentage": 75.31, "elapsed_time": "19:16:49", "remaining_time": "6:19:18"}
+{"current_steps": 675, "total_steps": 895, "loss": 0.6977, "lr": 1.3859237063453002e-05, "epoch": 3.752605976372481, "percentage": 75.42, "elapsed_time": "19:18:32", "remaining_time": "6:17:35"}
+{"current_steps": 676, "total_steps": 895, "loss": 0.7089, "lr": 1.3741279755644427e-05, "epoch": 3.7581653926337735, "percentage": 75.53, "elapsed_time": "19:20:13", "remaining_time": "6:15:52"}
+{"current_steps": 677, "total_steps": 895, "loss": 0.6907, "lr": 1.3623722375079345e-05, "epoch": 3.763724808895066, "percentage": 75.64, "elapsed_time": "19:21:55", "remaining_time": "6:14:09"}
+{"current_steps": 678, "total_steps": 895, "loss": 0.6964, "lr": 1.3506566712187703e-05, "epoch": 3.7692842251563587, "percentage": 75.75, "elapsed_time": "19:23:37", "remaining_time": "6:12:25"}
+{"current_steps": 679, "total_steps": 895, "loss": 0.7001, "lr": 1.338981455128117e-05, "epoch": 3.774843641417651, "percentage": 75.87, "elapsed_time": "19:25:19", "remaining_time": "6:10:42"}
+{"current_steps": 680, "total_steps": 895, "loss": 0.7137, "lr": 1.3273467670526018e-05, "epoch": 3.7804030576789436, "percentage": 75.98, "elapsed_time": "19:27:01", "remaining_time": "6:08:59"}
+{"current_steps": 681, "total_steps": 895, "loss": 0.6985, "lr": 1.3157527841915951e-05, "epoch": 3.7859624739402364, "percentage": 76.09, "elapsed_time": "19:28:42", "remaining_time": "6:07:15"}
+{"current_steps": 682, "total_steps": 895, "loss": 0.7127, "lr": 1.3041996831245172e-05, "epoch": 3.791521890201529, "percentage": 76.2, "elapsed_time": "19:30:25", "remaining_time": "6:05:32"}
+{"current_steps": 683, "total_steps": 895, "loss": 0.6918, "lr": 1.292687639808147e-05, "epoch": 3.7970813064628213, "percentage": 76.31, "elapsed_time": "19:32:09", "remaining_time": "6:03:49"}
+{"current_steps": 684, "total_steps": 895, "loss": 0.6926, "lr": 1.2812168295739466e-05, "epoch": 3.802640722724114, "percentage": 76.42, "elapsed_time": "19:33:52", "remaining_time": "6:02:06"}
+{"current_steps": 685, "total_steps": 895, "loss": 0.6878, "lr": 1.2697874271253844e-05, "epoch": 3.8082001389854065, "percentage": 76.54, "elapsed_time": "19:35:35", "remaining_time": "6:00:24"}
+{"current_steps": 686, "total_steps": 895, "loss": 0.7, "lr": 1.258399606535277e-05, "epoch": 3.813759555246699, "percentage": 76.65, "elapsed_time": "19:37:19", "remaining_time": "5:58:41"}
+{"current_steps": 687, "total_steps": 895, "loss": 0.6918, "lr": 1.2470535412431387e-05, "epoch": 3.819318971507992, "percentage": 76.76, "elapsed_time": "19:39:02", "remaining_time": "5:56:58"}
+{"current_steps": 688, "total_steps": 895, "loss": 0.709, "lr": 1.2357494040525416e-05, "epoch": 3.8248783877692842, "percentage": 76.87, "elapsed_time": "19:40:45", "remaining_time": "5:55:15"}
+{"current_steps": 689, "total_steps": 895, "loss": 0.7081, "lr": 1.2244873671284788e-05, "epoch": 3.8304378040305767, "percentage": 76.98, "elapsed_time": "19:42:26", "remaining_time": "5:53:31"}
+{"current_steps": 690, "total_steps": 895, "loss": 0.7073, "lr": 1.2132676019947493e-05, "epoch": 3.8359972202918695, "percentage": 77.09, "elapsed_time": "19:44:07", "remaining_time": "5:51:48"}
+{"current_steps": 691, "total_steps": 895, "loss": 0.6991, "lr": 1.2020902795313352e-05, "epoch": 3.841556636553162, "percentage": 77.21, "elapsed_time": "19:45:49", "remaining_time": "5:50:04"}
+{"current_steps": 692, "total_steps": 895, "loss": 0.6948, "lr": 1.1909555699718118e-05, "epoch": 3.8471160528144543, "percentage": 77.32, "elapsed_time": "19:47:31", "remaining_time": "5:48:21"}
+{"current_steps": 693, "total_steps": 895, "loss": 0.7176, "lr": 1.179863642900744e-05, "epoch": 3.852675469075747, "percentage": 77.43, "elapsed_time": "19:49:12", "remaining_time": "5:46:38"}
+{"current_steps": 694, "total_steps": 895, "loss": 0.7149, "lr": 1.1688146672511143e-05, "epoch": 3.8582348853370396, "percentage": 77.54, "elapsed_time": "19:50:55", "remaining_time": "5:44:55"}
+{"current_steps": 695, "total_steps": 895, "loss": 0.7015, "lr": 1.1578088113017337e-05, "epoch": 3.863794301598332, "percentage": 77.65, "elapsed_time": "19:52:38", "remaining_time": "5:43:12"}
+{"current_steps": 696, "total_steps": 895, "loss": 0.6942, "lr": 1.1468462426746988e-05, "epoch": 3.869353717859625, "percentage": 77.77, "elapsed_time": "19:54:21", "remaining_time": "5:41:29"}
+{"current_steps": 697, "total_steps": 895, "loss": 0.7117, "lr": 1.1359271283328224e-05, "epoch": 3.8749131341209173, "percentage": 77.88, "elapsed_time": "19:56:04", "remaining_time": "5:39:46"}
+{"current_steps": 698, "total_steps": 895, "loss": 0.7046, "lr": 1.1250516345771016e-05, "epoch": 3.8804725503822097, "percentage": 77.99, "elapsed_time": "19:57:46", "remaining_time": "5:38:03"}
+{"current_steps": 699, "total_steps": 895, "loss": 0.6949, "lr": 1.1142199270441774e-05, "epoch": 3.8860319666435026, "percentage": 78.1, "elapsed_time": "19:59:29", "remaining_time": "5:36:20"}
+{"current_steps": 700, "total_steps": 895, "loss": 0.7134, "lr": 1.1034321707038153e-05, "epoch": 3.891591382904795, "percentage": 78.21, "elapsed_time": "20:01:10", "remaining_time": "5:34:36"}
+{"current_steps": 701, "total_steps": 895, "loss": 0.6996, "lr": 1.092688529856393e-05, "epoch": 3.8971507991660874, "percentage": 78.32, "elapsed_time": "20:02:53", "remaining_time": "5:32:53"}
+{"current_steps": 702, "total_steps": 895, "loss": 0.712, "lr": 1.0819891681303973e-05, "epoch": 3.9027102154273803, "percentage": 78.44, "elapsed_time": "20:04:35", "remaining_time": "5:31:10"}
+{"current_steps": 703, "total_steps": 895, "loss": 0.7124, "lr": 1.0713342484799342e-05, "epoch": 3.9082696316886727, "percentage": 78.55, "elapsed_time": "20:06:17", "remaining_time": "5:29:27"}
+{"current_steps": 704, "total_steps": 895, "loss": 0.6946, "lr": 1.060723933182243e-05, "epoch": 3.913829047949965, "percentage": 78.66, "elapsed_time": "20:08:00", "remaining_time": "5:27:44"}
+{"current_steps": 705, "total_steps": 895, "loss": 0.7101, "lr": 1.0501583838352269e-05, "epoch": 3.919388464211258, "percentage": 78.77, "elapsed_time": "20:09:43", "remaining_time": "5:26:01"}
+{"current_steps": 706, "total_steps": 895, "loss": 0.6975, "lr": 1.0396377613549924e-05, "epoch": 3.9249478804725504, "percentage": 78.88, "elapsed_time": "20:11:26", "remaining_time": "5:24:18"}
+{"current_steps": 707, "total_steps": 895, "loss": 0.6898, "lr": 1.0291622259733995e-05, "epoch": 3.930507296733843, "percentage": 78.99, "elapsed_time": "20:13:09", "remaining_time": "5:22:35"}
+{"current_steps": 708, "total_steps": 895, "loss": 0.6942, "lr": 1.0187319372356175e-05, "epoch": 3.9360667129951357, "percentage": 79.11, "elapsed_time": "20:14:52", "remaining_time": "5:20:52"}
+{"current_steps": 709, "total_steps": 895, "loss": 0.7064, "lr": 1.0083470539976986e-05, "epoch": 3.941626129256428, "percentage": 79.22, "elapsed_time": "20:16:35", "remaining_time": "5:19:09"}
+{"current_steps": 710, "total_steps": 895, "loss": 0.7026, "lr": 9.980077344241547e-06, "epoch": 3.9471855455177205, "percentage": 79.33, "elapsed_time": "20:18:18", "remaining_time": "5:17:26"}
+{"current_steps": 711, "total_steps": 895, "loss": 0.7, "lr": 9.877141359855567e-06, "epoch": 3.9527449617790134, "percentage": 79.44, "elapsed_time": "20:20:02", "remaining_time": "5:15:44"}
+{"current_steps": 712, "total_steps": 895, "loss": 0.6884, "lr": 9.774664154561249e-06, "epoch": 3.9583043780403058, "percentage": 79.55, "elapsed_time": "20:21:46", "remaining_time": "5:14:01"}
+{"current_steps": 713, "total_steps": 895, "loss": 0.698, "lr": 9.672647289113497e-06, "epoch": 3.963863794301598, "percentage": 79.66, "elapsed_time": "20:23:27", "remaining_time": "5:12:18"}
+{"current_steps": 714, "total_steps": 895, "loss": 0.6899, "lr": 9.5710923172561e-06, "epoch": 3.969423210562891, "percentage": 79.78, "elapsed_time": "20:25:09", "remaining_time": "5:10:34"}
+{"current_steps": 715, "total_steps": 895, "loss": 0.7149, "lr": 9.470000785698117e-06, "epoch": 3.9749826268241835, "percentage": 79.89, "elapsed_time": "20:26:51", "remaining_time": "5:08:51"}
+{"current_steps": 716, "total_steps": 895, "loss": 0.6991, "lr": 9.369374234090248e-06, "epoch": 3.980542043085476, "percentage": 80.0, "elapsed_time": "20:28:34", "remaining_time": "5:07:08"}
+{"current_steps": 717, "total_steps": 895, "loss": 0.6986, "lr": 9.269214195001482e-06, "epoch": 3.9861014593467687, "percentage": 80.11, "elapsed_time": "20:30:15", "remaining_time": "5:05:25"}
+{"current_steps": 718, "total_steps": 895, "loss": 0.7133, "lr": 9.169522193895614e-06, "epoch": 3.991660875608061, "percentage": 80.22, "elapsed_time": "20:31:57", "remaining_time": "5:03:42"}
+{"current_steps": 719, "total_steps": 895, "loss": 0.6978, "lr": 9.070299749108184e-06, "epoch": 3.9972202918693536, "percentage": 80.34, "elapsed_time": "20:33:39", "remaining_time": "5:01:58"}
+{"current_steps": 720, "total_steps": 895, "loss": 0.6916, "lr": 8.971548371823205e-06, "epoch": 4.002779708130646, "percentage": 80.45, "elapsed_time": "20:37:10", "remaining_time": "5:00:42"}