Training in progress, epoch 3

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7918adc9b34addecc426ab0636e367447e07ff12f0bb6d6072078bf8c6b12453
+oid sha256:2a822226a28b6eb2ec9c9d7f726d8dc48eeeea69c83412fcfeff7c935ddfe7ca
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:532b6a17040d6bb502c60edd7e7d23632377d095b05221f693a6c313b837039c
+oid sha256:e9bd9bbe2669d7b8e4921d590087fb712d3a833623a36b2e3ab4a80dedfaba85
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4ab64792c2ee90ed85139a83b3e847413a74fd84398eb3fadbdcf9c05b4ab3a8
+oid sha256:f1e0d0e6cff60ecd7ba8d39fba6b132e66b4c243207269e571e6a1680cd21a4f
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ef0112349e5d19fed4f9b24c2e685036893c7d241481f1ed29f412d38a548f54
+oid sha256:d452653cf2452845ff5441d52d4f9d42c3561eb4e604a7f80fc37d9062c91353
 size 1089994880

trainer_log.jsonl

@@ -674,3 +674,183 @@
 {"current_steps": 541, "total_steps": 900, "loss": 0.7391, "lr": 3.29013420725106e-05, "epoch": 2.9937802349689013, "percentage": 60.11, "elapsed_time": "14:59:47", "remaining_time": "9:57:05"}
 {"current_steps": 542, "total_steps": 900, "loss": 0.7219, "lr": 3.274871803550834e-05, "epoch": 3.0006910850034556, "percentage": 60.22, "elapsed_time": "15:03:16", "remaining_time": "9:56:37"}
 {"current_steps": 543, "total_steps": 900, "loss": 0.7007, "lr": 3.2596203078211424e-05, "epoch": 3.0062197650310987, "percentage": 60.33, "elapsed_time": "15:04:55", "remaining_time": "9:54:57"}
+{"current_steps": 544, "total_steps": 900, "loss": 0.7142, "lr": 3.24437994948744e-05, "epoch": 3.011748445058742, "percentage": 60.44, "elapsed_time": "15:06:35", "remaining_time": "9:53:16"}
+{"current_steps": 545, "total_steps": 900, "loss": 0.6976, "lr": 3.229150957807641e-05, "epoch": 3.0172771250863857, "percentage": 60.56, "elapsed_time": "15:08:14", "remaining_time": "9:51:36"}
+{"current_steps": 546, "total_steps": 900, "loss": 0.6981, "lr": 3.213933561868679e-05, "epoch": 3.022805805114029, "percentage": 60.67, "elapsed_time": "15:09:53", "remaining_time": "9:49:55"}
+{"current_steps": 547, "total_steps": 900, "loss": 0.6937, "lr": 3.1987279905830505e-05, "epoch": 3.0283344851416722, "percentage": 60.78, "elapsed_time": "15:11:32", "remaining_time": "9:48:15"}
+{"current_steps": 548, "total_steps": 900, "loss": 0.6983, "lr": 3.183534472685378e-05, "epoch": 3.0338631651693158, "percentage": 60.89, "elapsed_time": "15:13:12", "remaining_time": "9:46:34"}
+{"current_steps": 549, "total_steps": 900, "loss": 0.7015, "lr": 3.168353236728964e-05, "epoch": 3.0393918451969593, "percentage": 61.0, "elapsed_time": "15:14:51", "remaining_time": "9:44:54"}
+{"current_steps": 550, "total_steps": 900, "loss": 0.6998, "lr": 3.153184511082359e-05, "epoch": 3.0449205252246028, "percentage": 61.11, "elapsed_time": "15:16:30", "remaining_time": "9:43:13"}
+{"current_steps": 551, "total_steps": 900, "loss": 0.7072, "lr": 3.138028523925921e-05, "epoch": 3.050449205252246, "percentage": 61.22, "elapsed_time": "15:18:09", "remaining_time": "9:41:33"}
+{"current_steps": 552, "total_steps": 900, "loss": 0.7171, "lr": 3.122885503248386e-05, "epoch": 3.0559778852798893, "percentage": 61.33, "elapsed_time": "15:19:49", "remaining_time": "9:39:53"}
+{"current_steps": 553, "total_steps": 900, "loss": 0.703, "lr": 3.107755676843443e-05, "epoch": 3.061506565307533, "percentage": 61.44, "elapsed_time": "15:21:28", "remaining_time": "9:38:12"}
+{"current_steps": 554, "total_steps": 900, "loss": 0.6908, "lr": 3.092639272306293e-05, "epoch": 3.0670352453351764, "percentage": 61.56, "elapsed_time": "15:23:07", "remaining_time": "9:36:32"}
+{"current_steps": 555, "total_steps": 900, "loss": 0.6912, "lr": 3.07753651703024e-05, "epoch": 3.0725639253628194, "percentage": 61.67, "elapsed_time": "15:24:47", "remaining_time": "9:34:51"}
+{"current_steps": 556, "total_steps": 900, "loss": 0.7021, "lr": 3.062447638203262e-05, "epoch": 3.078092605390463, "percentage": 61.78, "elapsed_time": "15:26:26", "remaining_time": "9:33:11"}
+{"current_steps": 557, "total_steps": 900, "loss": 0.7061, "lr": 3.0473728628045975e-05, "epoch": 3.0836212854181064, "percentage": 61.89, "elapsed_time": "15:28:05", "remaining_time": "9:31:31"}
+{"current_steps": 558, "total_steps": 900, "loss": 0.7005, "lr": 3.0323124176013297e-05, "epoch": 3.08914996544575, "percentage": 62.0, "elapsed_time": "15:29:45", "remaining_time": "9:29:50"}
+{"current_steps": 559, "total_steps": 900, "loss": 0.6969, "lr": 3.0172665291449747e-05, "epoch": 3.0946786454733934, "percentage": 62.11, "elapsed_time": "15:31:24", "remaining_time": "9:28:10"}
+{"current_steps": 560, "total_steps": 900, "loss": 0.6954, "lr": 3.0022354237680752e-05, "epoch": 3.1002073255010365, "percentage": 62.22, "elapsed_time": "15:33:03", "remaining_time": "9:26:30"}
+{"current_steps": 561, "total_steps": 900, "loss": 0.6912, "lr": 2.9872193275807933e-05, "epoch": 3.10573600552868, "percentage": 62.33, "elapsed_time": "15:34:42", "remaining_time": "9:24:49"}
+{"current_steps": 562, "total_steps": 900, "loss": 0.6838, "lr": 2.9722184664675123e-05, "epoch": 3.1112646855563235, "percentage": 62.44, "elapsed_time": "15:36:22", "remaining_time": "9:23:09"}
+{"current_steps": 563, "total_steps": 900, "loss": 0.6935, "lr": 2.9572330660834347e-05, "epoch": 3.116793365583967, "percentage": 62.56, "elapsed_time": "15:38:01", "remaining_time": "9:21:29"}
+{"current_steps": 564, "total_steps": 900, "loss": 0.701, "lr": 2.9422633518511926e-05, "epoch": 3.12232204561161, "percentage": 62.67, "elapsed_time": "15:39:41", "remaining_time": "9:19:49"}
+{"current_steps": 565, "total_steps": 900, "loss": 0.6942, "lr": 2.9273095489574502e-05, "epoch": 3.1278507256392536, "percentage": 62.78, "elapsed_time": "15:41:21", "remaining_time": "9:18:08"}
+{"current_steps": 566, "total_steps": 900, "loss": 0.7006, "lr": 2.9123718823495273e-05, "epoch": 3.133379405666897, "percentage": 62.89, "elapsed_time": "15:43:00", "remaining_time": "9:16:28"}
+{"current_steps": 567, "total_steps": 900, "loss": 0.7013, "lr": 2.8974505767320037e-05, "epoch": 3.1389080856945406, "percentage": 63.0, "elapsed_time": "15:44:40", "remaining_time": "9:14:48"}
+{"current_steps": 568, "total_steps": 900, "loss": 0.6959, "lr": 2.8825458565633432e-05, "epoch": 3.1444367657221837, "percentage": 63.11, "elapsed_time": "15:46:20", "remaining_time": "9:13:08"}
+{"current_steps": 569, "total_steps": 900, "loss": 0.6875, "lr": 2.8676579460525188e-05, "epoch": 3.149965445749827, "percentage": 63.22, "elapsed_time": "15:48:00", "remaining_time": "9:11:28"}
+{"current_steps": 570, "total_steps": 900, "loss": 0.6942, "lr": 2.8527870691556404e-05, "epoch": 3.1554941257774707, "percentage": 63.33, "elapsed_time": "15:49:39", "remaining_time": "9:09:48"}
+{"current_steps": 571, "total_steps": 900, "loss": 0.7143, "lr": 2.837933449572582e-05, "epoch": 3.161022805805114, "percentage": 63.44, "elapsed_time": "15:51:18", "remaining_time": "9:08:07"}
+{"current_steps": 572, "total_steps": 900, "loss": 0.6884, "lr": 2.8230973107436193e-05, "epoch": 3.1665514858327573, "percentage": 63.56, "elapsed_time": "15:52:58", "remaining_time": "9:06:27"}
+{"current_steps": 573, "total_steps": 900, "loss": 0.7133, "lr": 2.808278875846072e-05, "epoch": 3.1720801658604008, "percentage": 63.67, "elapsed_time": "15:54:37", "remaining_time": "9:04:47"}
+{"current_steps": 574, "total_steps": 900, "loss": 0.7034, "lr": 2.7934783677909384e-05, "epoch": 3.1776088458880443, "percentage": 63.78, "elapsed_time": "15:56:16", "remaining_time": "9:03:06"}
+{"current_steps": 575, "total_steps": 900, "loss": 0.6934, "lr": 2.778696009219548e-05, "epoch": 3.1831375259156878, "percentage": 63.89, "elapsed_time": "15:57:55", "remaining_time": "9:01:26"}
+{"current_steps": 576, "total_steps": 900, "loss": 0.6813, "lr": 2.7639320225002108e-05, "epoch": 3.188666205943331, "percentage": 64.0, "elapsed_time": "15:59:35", "remaining_time": "8:59:45"}
+{"current_steps": 577, "total_steps": 900, "loss": 0.699, "lr": 2.7491866297248727e-05, "epoch": 3.1941948859709743, "percentage": 64.11, "elapsed_time": "16:01:14", "remaining_time": "8:58:05"}
+{"current_steps": 578, "total_steps": 900, "loss": 0.707, "lr": 2.7344600527057736e-05, "epoch": 3.199723565998618, "percentage": 64.22, "elapsed_time": "16:02:53", "remaining_time": "8:56:25"}
+{"current_steps": 579, "total_steps": 900, "loss": 0.6982, "lr": 2.7197525129721138e-05, "epoch": 3.2052522460262614, "percentage": 64.33, "elapsed_time": "16:04:32", "remaining_time": "8:54:44"}
+{"current_steps": 580, "total_steps": 900, "loss": 0.6856, "lr": 2.7050642317667164e-05, "epoch": 3.2107809260539044, "percentage": 64.44, "elapsed_time": "16:06:12", "remaining_time": "8:53:04"}
+{"current_steps": 581, "total_steps": 900, "loss": 0.7004, "lr": 2.6903954300427027e-05, "epoch": 3.216309606081548, "percentage": 64.56, "elapsed_time": "16:07:51", "remaining_time": "8:51:24"}
+{"current_steps": 582, "total_steps": 900, "loss": 0.7, "lr": 2.6757463284601682e-05, "epoch": 3.2218382861091914, "percentage": 64.67, "elapsed_time": "16:09:30", "remaining_time": "8:49:44"}
+{"current_steps": 583, "total_steps": 900, "loss": 0.703, "lr": 2.6611171473828636e-05, "epoch": 3.227366966136835, "percentage": 64.78, "elapsed_time": "16:11:10", "remaining_time": "8:48:03"}
+{"current_steps": 584, "total_steps": 900, "loss": 0.6937, "lr": 2.646508106874876e-05, "epoch": 3.2328956461644784, "percentage": 64.89, "elapsed_time": "16:12:49", "remaining_time": "8:46:23"}
+{"current_steps": 585, "total_steps": 900, "loss": 0.7075, "lr": 2.6319194266973256e-05, "epoch": 3.2384243261921215, "percentage": 65.0, "elapsed_time": "16:14:28", "remaining_time": "8:44:43"}
+{"current_steps": 586, "total_steps": 900, "loss": 0.6998, "lr": 2.6173513263050535e-05, "epoch": 3.243953006219765, "percentage": 65.11, "elapsed_time": "16:16:07", "remaining_time": "8:43:02"}
+{"current_steps": 587, "total_steps": 900, "loss": 0.7098, "lr": 2.602804024843324e-05, "epoch": 3.2494816862474085, "percentage": 65.22, "elapsed_time": "16:17:47", "remaining_time": "8:41:22"}
+{"current_steps": 588, "total_steps": 900, "loss": 0.6888, "lr": 2.5882777411445254e-05, "epoch": 3.255010366275052, "percentage": 65.33, "elapsed_time": "16:19:26", "remaining_time": "8:39:42"}
+{"current_steps": 589, "total_steps": 900, "loss": 0.7032, "lr": 2.5737726937248804e-05, "epoch": 3.260539046302695, "percentage": 65.44, "elapsed_time": "16:21:05", "remaining_time": "8:38:01"}
+{"current_steps": 590, "total_steps": 900, "loss": 0.6995, "lr": 2.5592891007811594e-05, "epoch": 3.2660677263303386, "percentage": 65.56, "elapsed_time": "16:22:44", "remaining_time": "8:36:21"}
+{"current_steps": 591, "total_steps": 900, "loss": 0.7123, "lr": 2.5448271801873957e-05, "epoch": 3.271596406357982, "percentage": 65.67, "elapsed_time": "16:24:23", "remaining_time": "8:34:41"}
+{"current_steps": 592, "total_steps": 900, "loss": 0.696, "lr": 2.530387149491607e-05, "epoch": 3.2771250863856256, "percentage": 65.78, "elapsed_time": "16:26:03", "remaining_time": "8:33:00"}
+{"current_steps": 593, "total_steps": 900, "loss": 0.6966, "lr": 2.5159692259125326e-05, "epoch": 3.2826537664132687, "percentage": 65.89, "elapsed_time": "16:27:42", "remaining_time": "8:31:20"}
+{"current_steps": 594, "total_steps": 900, "loss": 0.7028, "lr": 2.501573626336352e-05, "epoch": 3.288182446440912, "percentage": 66.0, "elapsed_time": "16:29:21", "remaining_time": "8:29:40"}
+{"current_steps": 595, "total_steps": 900, "loss": 0.7066, "lr": 2.4872005673134307e-05, "epoch": 3.2937111264685557, "percentage": 66.11, "elapsed_time": "16:31:00", "remaining_time": "8:27:59"}
+{"current_steps": 596, "total_steps": 900, "loss": 0.6979, "lr": 2.47285026505506e-05, "epoch": 3.299239806496199, "percentage": 66.22, "elapsed_time": "16:32:40", "remaining_time": "8:26:19"}
+{"current_steps": 597, "total_steps": 900, "loss": 0.7049, "lr": 2.4585229354302077e-05, "epoch": 3.3047684865238423, "percentage": 66.33, "elapsed_time": "16:34:19", "remaining_time": "8:24:39"}
+{"current_steps": 598, "total_steps": 900, "loss": 0.6993, "lr": 2.4442187939622643e-05, "epoch": 3.3102971665514858, "percentage": 66.44, "elapsed_time": "16:35:58", "remaining_time": "8:22:59"}
+{"current_steps": 599, "total_steps": 900, "loss": 0.6844, "lr": 2.4299380558258116e-05, "epoch": 3.3158258465791293, "percentage": 66.56, "elapsed_time": "16:37:38", "remaining_time": "8:21:18"}
+{"current_steps": 600, "total_steps": 900, "loss": 0.714, "lr": 2.4156809358433728e-05, "epoch": 3.3213545266067728, "percentage": 66.67, "elapsed_time": "16:39:17", "remaining_time": "8:19:38"}
+{"current_steps": 601, "total_steps": 900, "loss": 0.7025, "lr": 2.401447648482192e-05, "epoch": 3.326883206634416, "percentage": 66.78, "elapsed_time": "16:40:56", "remaining_time": "8:17:58"}
+{"current_steps": 602, "total_steps": 900, "loss": 0.71, "lr": 2.3872384078509998e-05, "epoch": 3.3324118866620593, "percentage": 66.89, "elapsed_time": "16:42:35", "remaining_time": "8:16:18"}
+{"current_steps": 603, "total_steps": 900, "loss": 0.6989, "lr": 2.3730534276968e-05, "epoch": 3.337940566689703, "percentage": 67.0, "elapsed_time": "16:44:15", "remaining_time": "8:14:37"}
+{"current_steps": 604, "total_steps": 900, "loss": 0.6879, "lr": 2.358892921401648e-05, "epoch": 3.3434692467173464, "percentage": 67.11, "elapsed_time": "16:45:54", "remaining_time": "8:12:57"}
+{"current_steps": 605, "total_steps": 900, "loss": 0.7079, "lr": 2.3447571019794438e-05, "epoch": 3.34899792674499, "percentage": 67.22, "elapsed_time": "16:47:33", "remaining_time": "8:11:17"}
+{"current_steps": 606, "total_steps": 900, "loss": 0.6989, "lr": 2.330646182072729e-05, "epoch": 3.354526606772633, "percentage": 67.33, "elapsed_time": "16:49:12", "remaining_time": "8:09:37"}
+{"current_steps": 607, "total_steps": 900, "loss": 0.6962, "lr": 2.316560373949484e-05, "epoch": 3.3600552868002764, "percentage": 67.44, "elapsed_time": "16:50:52", "remaining_time": "8:07:56"}
+{"current_steps": 608, "total_steps": 900, "loss": 0.7048, "lr": 2.3024998894999388e-05, "epoch": 3.36558396682792, "percentage": 67.56, "elapsed_time": "16:52:31", "remaining_time": "8:06:16"}
+{"current_steps": 609, "total_steps": 900, "loss": 0.6974, "lr": 2.288464940233384e-05, "epoch": 3.371112646855563, "percentage": 67.67, "elapsed_time": "16:54:10", "remaining_time": "8:04:36"}
+{"current_steps": 610, "total_steps": 900, "loss": 0.697, "lr": 2.274455737274987e-05, "epoch": 3.3766413268832065, "percentage": 67.78, "elapsed_time": "16:55:49", "remaining_time": "8:02:56"}
+{"current_steps": 611, "total_steps": 900, "loss": 0.6958, "lr": 2.2604724913626216e-05, "epoch": 3.38217000691085, "percentage": 67.89, "elapsed_time": "16:57:29", "remaining_time": "8:01:15"}
+{"current_steps": 612, "total_steps": 900, "loss": 0.6997, "lr": 2.24651541284369e-05, "epoch": 3.3876986869384935, "percentage": 68.0, "elapsed_time": "16:59:08", "remaining_time": "7:59:35"}
+{"current_steps": 613, "total_steps": 900, "loss": 0.7027, "lr": 2.2325847116719698e-05, "epoch": 3.393227366966137, "percentage": 68.11, "elapsed_time": "17:00:47", "remaining_time": "7:57:55"}
+{"current_steps": 614, "total_steps": 900, "loss": 0.7084, "lr": 2.2186805974044426e-05, "epoch": 3.39875604699378, "percentage": 68.22, "elapsed_time": "17:02:26", "remaining_time": "7:56:15"}
+{"current_steps": 615, "total_steps": 900, "loss": 0.7063, "lr": 2.2048032791981515e-05, "epoch": 3.4042847270214236, "percentage": 68.33, "elapsed_time": "17:04:06", "remaining_time": "7:54:35"}
+{"current_steps": 616, "total_steps": 900, "loss": 0.7025, "lr": 2.19095296580705e-05, "epoch": 3.409813407049067, "percentage": 68.44, "elapsed_time": "17:05:45", "remaining_time": "7:52:54"}
+{"current_steps": 617, "total_steps": 900, "loss": 0.6867, "lr": 2.177129865578863e-05, "epoch": 3.4153420870767106, "percentage": 68.56, "elapsed_time": "17:07:24", "remaining_time": "7:51:14"}
+{"current_steps": 618, "total_steps": 900, "loss": 0.6904, "lr": 2.1633341864519526e-05, "epoch": 3.4208707671043537, "percentage": 68.67, "elapsed_time": "17:09:03", "remaining_time": "7:49:34"}
+{"current_steps": 619, "total_steps": 900, "loss": 0.695, "lr": 2.1495661359521937e-05, "epoch": 3.426399447131997, "percentage": 68.78, "elapsed_time": "17:10:43", "remaining_time": "7:47:54"}
+{"current_steps": 620, "total_steps": 900, "loss": 0.695, "lr": 2.135825921189846e-05, "epoch": 3.4319281271596407, "percentage": 68.89, "elapsed_time": "17:12:22", "remaining_time": "7:46:14"}
+{"current_steps": 621, "total_steps": 900, "loss": 0.6924, "lr": 2.122113748856438e-05, "epoch": 3.437456807187284, "percentage": 69.0, "elapsed_time": "17:14:01", "remaining_time": "7:44:33"}
+{"current_steps": 622, "total_steps": 900, "loss": 0.6956, "lr": 2.108429825221667e-05, "epoch": 3.4429854872149273, "percentage": 69.11, "elapsed_time": "17:15:40", "remaining_time": "7:42:53"}
+{"current_steps": 623, "total_steps": 900, "loss": 0.6862, "lr": 2.094774356130287e-05, "epoch": 3.4485141672425708, "percentage": 69.22, "elapsed_time": "17:17:20", "remaining_time": "7:41:13"}
+{"current_steps": 624, "total_steps": 900, "loss": 0.7051, "lr": 2.0811475469990167e-05, "epoch": 3.4540428472702143, "percentage": 69.33, "elapsed_time": "17:18:59", "remaining_time": "7:39:33"}
+{"current_steps": 625, "total_steps": 900, "loss": 0.7036, "lr": 2.067549602813446e-05, "epoch": 3.4595715272978578, "percentage": 69.44, "elapsed_time": "17:20:38", "remaining_time": "7:37:53"}
+{"current_steps": 626, "total_steps": 900, "loss": 0.6989, "lr": 2.0539807281249608e-05, "epoch": 3.4651002073255013, "percentage": 69.56, "elapsed_time": "17:22:17", "remaining_time": "7:36:12"}
+{"current_steps": 627, "total_steps": 900, "loss": 0.7093, "lr": 2.0404411270476527e-05, "epoch": 3.4706288873531443, "percentage": 69.67, "elapsed_time": "17:23:57", "remaining_time": "7:34:32"}
+{"current_steps": 628, "total_steps": 900, "loss": 0.7088, "lr": 2.026931003255259e-05, "epoch": 3.476157567380788, "percentage": 69.78, "elapsed_time": "17:25:36", "remaining_time": "7:32:52"}
+{"current_steps": 629, "total_steps": 900, "loss": 0.7095, "lr": 2.0134505599780962e-05, "epoch": 3.4816862474084314, "percentage": 69.89, "elapsed_time": "17:27:15", "remaining_time": "7:31:12"}
+{"current_steps": 630, "total_steps": 900, "loss": 0.6992, "lr": 2.0000000000000012e-05, "epoch": 3.4872149274360744, "percentage": 70.0, "elapsed_time": "17:28:55", "remaining_time": "7:29:32"}
+{"current_steps": 631, "total_steps": 900, "loss": 0.7017, "lr": 1.986579525655281e-05, "epoch": 3.492743607463718, "percentage": 70.11, "elapsed_time": "17:30:34", "remaining_time": "7:27:51"}
+{"current_steps": 632, "total_steps": 900, "loss": 0.6954, "lr": 1.973189338825671e-05, "epoch": 3.4982722874913614, "percentage": 70.22, "elapsed_time": "17:32:13", "remaining_time": "7:26:11"}
+{"current_steps": 633, "total_steps": 900, "loss": 0.7034, "lr": 1.959829640937299e-05, "epoch": 3.503800967519005, "percentage": 70.33, "elapsed_time": "17:33:52", "remaining_time": "7:24:31"}
+{"current_steps": 634, "total_steps": 900, "loss": 0.7105, "lr": 1.9465006329576502e-05, "epoch": 3.5093296475466484, "percentage": 70.44, "elapsed_time": "17:35:31", "remaining_time": "7:22:51"}
+{"current_steps": 635, "total_steps": 900, "loss": 0.6992, "lr": 1.9332025153925486e-05, "epoch": 3.5148583275742915, "percentage": 70.56, "elapsed_time": "17:37:11", "remaining_time": "7:21:11"}
+{"current_steps": 636, "total_steps": 900, "loss": 0.6964, "lr": 1.9199354882831387e-05, "epoch": 3.520387007601935, "percentage": 70.67, "elapsed_time": "17:38:50", "remaining_time": "7:19:31"}
+{"current_steps": 637, "total_steps": 900, "loss": 0.6826, "lr": 1.906699751202877e-05, "epoch": 3.5259156876295785, "percentage": 70.78, "elapsed_time": "17:40:29", "remaining_time": "7:17:51"}
+{"current_steps": 638, "total_steps": 900, "loss": 0.7029, "lr": 1.8934955032545307e-05, "epoch": 3.5314443676572216, "percentage": 70.89, "elapsed_time": "17:42:08", "remaining_time": "7:16:10"}
+{"current_steps": 639, "total_steps": 900, "loss": 0.703, "lr": 1.880322943067181e-05, "epoch": 3.536973047684865, "percentage": 71.0, "elapsed_time": "17:43:48", "remaining_time": "7:14:30"}
+{"current_steps": 640, "total_steps": 900, "loss": 0.7079, "lr": 1.867182268793236e-05, "epoch": 3.5425017277125086, "percentage": 71.11, "elapsed_time": "17:45:27", "remaining_time": "7:12:50"}
+{"current_steps": 641, "total_steps": 900, "loss": 0.6974, "lr": 1.8540736781054497e-05, "epoch": 3.548030407740152, "percentage": 71.22, "elapsed_time": "17:47:06", "remaining_time": "7:11:10"}
+{"current_steps": 642, "total_steps": 900, "loss": 0.6991, "lr": 1.8409973681939498e-05, "epoch": 3.5535590877677956, "percentage": 71.33, "elapsed_time": "17:48:45", "remaining_time": "7:09:30"}
+{"current_steps": 643, "total_steps": 900, "loss": 0.7071, "lr": 1.827953535763268e-05, "epoch": 3.5590877677954387, "percentage": 71.44, "elapsed_time": "17:50:25", "remaining_time": "7:07:50"}
+{"current_steps": 644, "total_steps": 900, "loss": 0.6994, "lr": 1.8149423770293855e-05, "epoch": 3.564616447823082, "percentage": 71.56, "elapsed_time": "17:52:04", "remaining_time": "7:06:09"}
+{"current_steps": 645, "total_steps": 900, "loss": 0.7022, "lr": 1.8019640877167763e-05, "epoch": 3.5701451278507257, "percentage": 71.67, "elapsed_time": "17:53:43", "remaining_time": "7:04:29"}
+{"current_steps": 646, "total_steps": 900, "loss": 0.7011, "lr": 1.7890188630554695e-05, "epoch": 3.575673807878369, "percentage": 71.78, "elapsed_time": "17:55:22", "remaining_time": "7:02:49"}
+{"current_steps": 647, "total_steps": 900, "loss": 0.7043, "lr": 1.7761068977781066e-05, "epoch": 3.5812024879060127, "percentage": 71.89, "elapsed_time": "17:57:02", "remaining_time": "7:01:09"}
+{"current_steps": 648, "total_steps": 900, "loss": 0.6993, "lr": 1.7632283861170135e-05, "epoch": 3.5867311679336558, "percentage": 72.0, "elapsed_time": "17:58:41", "remaining_time": "6:59:29"}
+{"current_steps": 649, "total_steps": 900, "loss": 0.6927, "lr": 1.7503835218012805e-05, "epoch": 3.5922598479612993, "percentage": 72.11, "elapsed_time": "18:00:20", "remaining_time": "6:57:49"}
+{"current_steps": 650, "total_steps": 900, "loss": 0.6913, "lr": 1.7375724980538465e-05, "epoch": 3.5977885279889428, "percentage": 72.22, "elapsed_time": "18:01:59", "remaining_time": "6:56:09"}
+{"current_steps": 651, "total_steps": 900, "loss": 0.6963, "lr": 1.7247955075885938e-05, "epoch": 3.603317208016586, "percentage": 72.33, "elapsed_time": "18:03:39", "remaining_time": "6:54:29"}
+{"current_steps": 652, "total_steps": 900, "loss": 0.6936, "lr": 1.7120527426074466e-05, "epoch": 3.6088458880442293, "percentage": 72.44, "elapsed_time": "18:05:18", "remaining_time": "6:52:48"}
+{"current_steps": 653, "total_steps": 900, "loss": 0.7066, "lr": 1.6993443947974846e-05, "epoch": 3.614374568071873, "percentage": 72.56, "elapsed_time": "18:06:57", "remaining_time": "6:51:08"}
+{"current_steps": 654, "total_steps": 900, "loss": 0.7029, "lr": 1.686670655328054e-05, "epoch": 3.6199032480995164, "percentage": 72.67, "elapsed_time": "18:08:36", "remaining_time": "6:49:28"}
+{"current_steps": 655, "total_steps": 900, "loss": 0.7034, "lr": 1.6740317148478932e-05, "epoch": 3.62543192812716, "percentage": 72.78, "elapsed_time": "18:10:16", "remaining_time": "6:47:48"}
+{"current_steps": 656, "total_steps": 900, "loss": 0.7028, "lr": 1.661427763482268e-05, "epoch": 3.630960608154803, "percentage": 72.89, "elapsed_time": "18:11:55", "remaining_time": "6:46:08"}
+{"current_steps": 657, "total_steps": 900, "loss": 0.7101, "lr": 1.648858990830108e-05, "epoch": 3.6364892881824464, "percentage": 73.0, "elapsed_time": "18:13:34", "remaining_time": "6:44:28"}
+{"current_steps": 658, "total_steps": 900, "loss": 0.6873, "lr": 1.6363255859611556e-05, "epoch": 3.64201796821009, "percentage": 73.11, "elapsed_time": "18:15:14", "remaining_time": "6:42:48"}
+{"current_steps": 659, "total_steps": 900, "loss": 0.7052, "lr": 1.6238277374131235e-05, "epoch": 3.647546648237733, "percentage": 73.22, "elapsed_time": "18:16:53", "remaining_time": "6:41:08"}
+{"current_steps": 660, "total_steps": 900, "loss": 0.7029, "lr": 1.6113656331888563e-05, "epoch": 3.6530753282653765, "percentage": 73.33, "elapsed_time": "18:18:32", "remaining_time": "6:39:28"}
+{"current_steps": 661, "total_steps": 900, "loss": 0.6905, "lr": 1.598939460753503e-05, "epoch": 3.65860400829302, "percentage": 73.44, "elapsed_time": "18:20:11", "remaining_time": "6:37:48"}
+{"current_steps": 662, "total_steps": 900, "loss": 0.7001, "lr": 1.5865494070316975e-05, "epoch": 3.6641326883206635, "percentage": 73.56, "elapsed_time": "18:21:50", "remaining_time": "6:36:07"}
+{"current_steps": 663, "total_steps": 900, "loss": 0.695, "lr": 1.5741956584047478e-05, "epoch": 3.669661368348307, "percentage": 73.67, "elapsed_time": "18:23:30", "remaining_time": "6:34:27"}
+{"current_steps": 664, "total_steps": 900, "loss": 0.6943, "lr": 1.5618784007078285e-05, "epoch": 3.67519004837595, "percentage": 73.78, "elapsed_time": "18:25:09", "remaining_time": "6:32:47"}
+{"current_steps": 665, "total_steps": 900, "loss": 0.6896, "lr": 1.5495978192271887e-05, "epoch": 3.6807187284035936, "percentage": 73.89, "elapsed_time": "18:26:48", "remaining_time": "6:31:07"}
+{"current_steps": 666, "total_steps": 900, "loss": 0.7166, "lr": 1.537354098697367e-05, "epoch": 3.686247408431237, "percentage": 74.0, "elapsed_time": "18:28:27", "remaining_time": "6:29:27"}
+{"current_steps": 667, "total_steps": 900, "loss": 0.7025, "lr": 1.5251474232984058e-05, "epoch": 3.69177608845888, "percentage": 74.11, "elapsed_time": "18:30:07", "remaining_time": "6:27:47"}
+{"current_steps": 668, "total_steps": 900, "loss": 0.7132, "lr": 1.5129779766530849e-05, "epoch": 3.697304768486524, "percentage": 74.22, "elapsed_time": "18:31:46", "remaining_time": "6:26:07"}
+{"current_steps": 669, "total_steps": 900, "loss": 0.6998, "lr": 1.5008459418241601e-05, "epoch": 3.702833448514167, "percentage": 74.33, "elapsed_time": "18:33:25", "remaining_time": "6:24:27"}
+{"current_steps": 670, "total_steps": 900, "loss": 0.6861, "lr": 1.4887515013116067e-05, "epoch": 3.7083621285418107, "percentage": 74.44, "elapsed_time": "18:35:04", "remaining_time": "6:22:47"}
+{"current_steps": 671, "total_steps": 900, "loss": 0.693, "lr": 1.476694837049875e-05, "epoch": 3.713890808569454, "percentage": 74.56, "elapsed_time": "18:36:44", "remaining_time": "6:21:07"}
+{"current_steps": 672, "total_steps": 900, "loss": 0.7068, "lr": 1.4646761304051587e-05, "epoch": 3.7194194885970973, "percentage": 74.67, "elapsed_time": "18:38:23", "remaining_time": "6:19:27"}
+{"current_steps": 673, "total_steps": 900, "loss": 0.7106, "lr": 1.4526955621726564e-05, "epoch": 3.7249481686247408, "percentage": 74.78, "elapsed_time": "18:40:02", "remaining_time": "6:17:47"}
+{"current_steps": 674, "total_steps": 900, "loss": 0.708, "lr": 1.4407533125738606e-05, "epoch": 3.7304768486523843, "percentage": 74.89, "elapsed_time": "18:41:41", "remaining_time": "6:16:07"}
+{"current_steps": 675, "total_steps": 900, "loss": 0.6953, "lr": 1.4288495612538427e-05, "epoch": 3.7360055286800278, "percentage": 75.0, "elapsed_time": "18:43:21", "remaining_time": "6:14:27"}
+{"current_steps": 676, "total_steps": 900, "loss": 0.6972, "lr": 1.4169844872785517e-05, "epoch": 3.7415342087076713, "percentage": 75.11, "elapsed_time": "18:45:00", "remaining_time": "6:12:46"}
+{"current_steps": 677, "total_steps": 900, "loss": 0.6874, "lr": 1.40515826913212e-05, "epoch": 3.7470628887353143, "percentage": 75.22, "elapsed_time": "18:46:39", "remaining_time": "6:11:06"}
+{"current_steps": 678, "total_steps": 900, "loss": 0.6915, "lr": 1.3933710847141795e-05, "epoch": 3.752591568762958, "percentage": 75.33, "elapsed_time": "18:48:19", "remaining_time": "6:09:26"}
+{"current_steps": 679, "total_steps": 900, "loss": 0.6998, "lr": 1.3816231113371851e-05, "epoch": 3.7581202487906014, "percentage": 75.44, "elapsed_time": "18:49:58", "remaining_time": "6:07:46"}
+{"current_steps": 680, "total_steps": 900, "loss": 0.7012, "lr": 1.369914525723746e-05, "epoch": 3.7636489288182444, "percentage": 75.56, "elapsed_time": "18:51:37", "remaining_time": "6:06:06"}
+{"current_steps": 681, "total_steps": 900, "loss": 0.7132, "lr": 1.3582455040039699e-05, "epoch": 3.769177608845888, "percentage": 75.67, "elapsed_time": "18:53:16", "remaining_time": "6:04:26"}
+{"current_steps": 682, "total_steps": 900, "loss": 0.6938, "lr": 1.3466162217128113e-05, "epoch": 3.7747062888735314, "percentage": 75.78, "elapsed_time": "18:54:55", "remaining_time": "6:02:46"}
+{"current_steps": 683, "total_steps": 900, "loss": 0.6801, "lr": 1.3350268537874316e-05, "epoch": 3.780234968901175, "percentage": 75.89, "elapsed_time": "18:56:35", "remaining_time": "6:01:06"}
+{"current_steps": 684, "total_steps": 900, "loss": 0.7054, "lr": 1.3234775745645684e-05, "epoch": 3.7857636489288184, "percentage": 76.0, "elapsed_time": "18:58:14", "remaining_time": "5:59:26"}
+{"current_steps": 685, "total_steps": 900, "loss": 0.6999, "lr": 1.3119685577779105e-05, "epoch": 3.7912923289564615, "percentage": 76.11, "elapsed_time": "18:59:53", "remaining_time": "5:57:46"}
+{"current_steps": 686, "total_steps": 900, "loss": 0.7217, "lr": 1.3004999765554902e-05, "epoch": 3.796821008984105, "percentage": 76.22, "elapsed_time": "19:01:33", "remaining_time": "5:56:06"}
+{"current_steps": 687, "total_steps": 900, "loss": 0.692, "lr": 1.2890720034170712e-05, "epoch": 3.8023496890117485, "percentage": 76.33, "elapsed_time": "19:03:12", "remaining_time": "5:54:26"}
+{"current_steps": 688, "total_steps": 900, "loss": 0.7077, "lr": 1.2776848102715587e-05, "epoch": 3.8078783690393916, "percentage": 76.44, "elapsed_time": "19:04:51", "remaining_time": "5:52:46"}
+{"current_steps": 689, "total_steps": 900, "loss": 0.7129, "lr": 1.2663385684144114e-05, "epoch": 3.813407049067035, "percentage": 76.56, "elapsed_time": "19:06:30", "remaining_time": "5:51:06"}
+{"current_steps": 690, "total_steps": 900, "loss": 0.7102, "lr": 1.2550334485250661e-05, "epoch": 3.8189357290946786, "percentage": 76.67, "elapsed_time": "19:08:10", "remaining_time": "5:49:26"}
+{"current_steps": 691, "total_steps": 900, "loss": 0.6974, "lr": 1.243769620664368e-05, "epoch": 3.824464409122322, "percentage": 76.78, "elapsed_time": "19:09:49", "remaining_time": "5:47:46"}
+{"current_steps": 692, "total_steps": 900, "loss": 0.6898, "lr": 1.2325472542720163e-05, "epoch": 3.8299930891499656, "percentage": 76.89, "elapsed_time": "19:11:28", "remaining_time": "5:46:06"}
+{"current_steps": 693, "total_steps": 900, "loss": 0.6981, "lr": 1.2213665181640106e-05, "epoch": 3.8355217691776087, "percentage": 77.0, "elapsed_time": "19:13:07", "remaining_time": "5:44:26"}
+{"current_steps": 694, "total_steps": 900, "loss": 0.6862, "lr": 1.210227580530115e-05, "epoch": 3.841050449205252, "percentage": 77.11, "elapsed_time": "19:14:46", "remaining_time": "5:42:46"}
+{"current_steps": 695, "total_steps": 900, "loss": 0.6938, "lr": 1.1991306089313261e-05, "epoch": 3.8465791292328957, "percentage": 77.22, "elapsed_time": "19:16:26", "remaining_time": "5:41:06"}
+{"current_steps": 696, "total_steps": 900, "loss": 0.6837, "lr": 1.1880757702973531e-05, "epoch": 3.852107809260539, "percentage": 77.33, "elapsed_time": "19:18:05", "remaining_time": "5:39:26"}
+{"current_steps": 697, "total_steps": 900, "loss": 0.6991, "lr": 1.177063230924108e-05, "epoch": 3.8576364892881827, "percentage": 77.44, "elapsed_time": "19:19:44", "remaining_time": "5:37:46"}
+{"current_steps": 698, "total_steps": 900, "loss": 0.7083, "lr": 1.1660931564712011e-05, "epoch": 3.8631651693158258, "percentage": 77.56, "elapsed_time": "19:21:24", "remaining_time": "5:36:06"}
+{"current_steps": 699, "total_steps": 900, "loss": 0.709, "lr": 1.1551657119594517e-05, "epoch": 3.8686938493434693, "percentage": 77.67, "elapsed_time": "19:23:03", "remaining_time": "5:34:26"}
+{"current_steps": 700, "total_steps": 900, "loss": 0.6993, "lr": 1.1442810617684046e-05, "epoch": 3.8742225293711128, "percentage": 77.78, "elapsed_time": "19:24:42", "remaining_time": "5:32:46"}
+{"current_steps": 701, "total_steps": 900, "loss": 0.689, "lr": 1.133439369633857e-05, "epoch": 3.879751209398756, "percentage": 77.89, "elapsed_time": "19:26:22", "remaining_time": "5:31:06"}
+{"current_steps": 702, "total_steps": 900, "loss": 0.6973, "lr": 1.1226407986453963e-05, "epoch": 3.8852798894263993, "percentage": 78.0, "elapsed_time": "19:28:01", "remaining_time": "5:29:26"}
+{"current_steps": 703, "total_steps": 900, "loss": 0.6872, "lr": 1.1118855112439468e-05, "epoch": 3.890808569454043, "percentage": 78.11, "elapsed_time": "19:29:40", "remaining_time": "5:27:46"}
+{"current_steps": 704, "total_steps": 900, "loss": 0.7097, "lr": 1.1011736692193246e-05, "epoch": 3.8963372494816864, "percentage": 78.22, "elapsed_time": "19:31:19", "remaining_time": "5:26:06"}
+{"current_steps": 705, "total_steps": 900, "loss": 0.7039, "lr": 1.0905054337078051e-05, "epoch": 3.90186592950933, "percentage": 78.33, "elapsed_time": "19:32:58", "remaining_time": "5:24:26"}
+{"current_steps": 706, "total_steps": 900, "loss": 0.7045, "lr": 1.0798809651897017e-05, "epoch": 3.907394609536973, "percentage": 78.44, "elapsed_time": "19:34:38", "remaining_time": "5:22:46"}
+{"current_steps": 707, "total_steps": 900, "loss": 0.692, "lr": 1.0693004234869458e-05, "epoch": 3.9129232895646164, "percentage": 78.56, "elapsed_time": "19:36:17", "remaining_time": "5:21:06"}
+{"current_steps": 708, "total_steps": 900, "loss": 0.7035, "lr": 1.0587639677606857e-05, "epoch": 3.91845196959226, "percentage": 78.67, "elapsed_time": "19:37:56", "remaining_time": "5:19:26"}
+{"current_steps": 709, "total_steps": 900, "loss": 0.6996, "lr": 1.0482717565088944e-05, "epoch": 3.923980649619903, "percentage": 78.78, "elapsed_time": "19:39:35", "remaining_time": "5:17:46"}
+{"current_steps": 710, "total_steps": 900, "loss": 0.6874, "lr": 1.0378239475639823e-05, "epoch": 3.9295093296475465, "percentage": 78.89, "elapsed_time": "19:41:15", "remaining_time": "5:16:06"}
+{"current_steps": 711, "total_steps": 900, "loss": 0.6952, "lr": 1.0274206980904226e-05, "epoch": 3.93503800967519, "percentage": 79.0, "elapsed_time": "19:42:54", "remaining_time": "5:14:26"}
+{"current_steps": 712, "total_steps": 900, "loss": 0.6949, "lr": 1.0170621645823928e-05, "epoch": 3.9405666897028335, "percentage": 79.11, "elapsed_time": "19:44:34", "remaining_time": "5:12:46"}
+{"current_steps": 713, "total_steps": 900, "loss": 0.7061, "lr": 1.0067485028614122e-05, "epoch": 3.946095369730477, "percentage": 79.22, "elapsed_time": "19:46:13", "remaining_time": "5:11:06"}
+{"current_steps": 714, "total_steps": 900, "loss": 0.7042, "lr": 9.964798680740033e-06, "epoch": 3.95162404975812, "percentage": 79.33, "elapsed_time": "19:47:53", "remaining_time": "5:09:26"}
+{"current_steps": 715, "total_steps": 900, "loss": 0.7002, "lr": 9.862564146893571e-06, "epoch": 3.9571527297857636, "percentage": 79.44, "elapsed_time": "19:49:32", "remaining_time": "5:07:47"}
+{"current_steps": 716, "total_steps": 900, "loss": 0.7097, "lr": 9.760782964970085e-06, "epoch": 3.962681409813407, "percentage": 79.56, "elapsed_time": "19:51:12", "remaining_time": "5:06:07"}
+{"current_steps": 717, "total_steps": 900, "loss": 0.7005, "lr": 9.659456666045247e-06, "epoch": 3.9682100898410506, "percentage": 79.67, "elapsed_time": "19:52:52", "remaining_time": "5:04:27"}
+{"current_steps": 718, "total_steps": 900, "loss": 0.6905, "lr": 9.558586774351983e-06, "epoch": 3.973738769868694, "percentage": 79.78, "elapsed_time": "19:54:32", "remaining_time": "5:02:47"}
+{"current_steps": 719, "total_steps": 900, "loss": 0.6804, "lr": 9.458174807257592e-06, "epoch": 3.979267449896337, "percentage": 79.89, "elapsed_time": "19:56:11", "remaining_time": "5:01:07"}
+{"current_steps": 720, "total_steps": 900, "loss": 0.7063, "lr": 9.358222275240884e-06, "epoch": 3.9847961299239807, "percentage": 80.0, "elapsed_time": "19:57:50", "remaining_time": "4:59:27"}
+{"current_steps": 721, "total_steps": 900, "loss": 0.6893, "lr": 9.258730681869478e-06, "epoch": 3.990324809951624, "percentage": 80.11, "elapsed_time": "19:59:29", "remaining_time": "4:57:47"}
+{"current_steps": 722, "total_steps": 900, "loss": 0.7101, "lr": 9.159701523777178e-06, "epoch": 3.9958534899792673, "percentage": 80.22, "elapsed_time": "20:01:06", "remaining_time": "4:56:07"}
+{"current_steps": 723, "total_steps": 900, "loss": 0.6886, "lr": 9.061136290641448e-06, "epoch": 4.0027643400138215, "percentage": 80.33, "elapsed_time": "20:04:33", "remaining_time": "4:54:53"}