Training in progress, epoch 4

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:691f8b2a09b5667d973061316aed2a701bee1d5175c0a3e504ceb1c0f2e54693
+oid sha256:ce4cdaee121e87216906a4af2739c5d139ccd6bb4addeaec068c95c0457c1451
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:199c1dd7637999046463543c934ac4c4040aa34404d5a39bd6f026d7992093fc
+oid sha256:a28efc685e82d9d611d5469f5cc39b7424bd1c0328f391d3c28dbf6ea424fd85
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c43e85639b1b279b1c3a5bf731e1dec44c6bd151f1946d4c67e4a273e0c98648
+oid sha256:0ee4e14a85a2b3120143ad2de92687e81c6b534eb0fb4a1fe3cb6c61326a64e4
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d40d515a91368d38b52ed5c9364c6fb9cd6c4a3e68e909c37357545ee61d2792
+oid sha256:38962cbaaccea53ada55149d2729a5bf819e142f94331c83199239f28e729fd9
 size 1089994880

trainer_log.jsonl

@@ -549,3 +549,187 @@
 {"current_steps": 549, "total_steps": 915, "loss": 0.8936, "lr": 3.3087490966086075e-05, "epoch": 2.9959072305593453, "percentage": 60.0, "elapsed_time": "8:08:18", "remaining_time": "5:25:32"}
 {"current_steps": 550, "total_steps": 915, "loss": 1.5762, "lr": 3.293714915649045e-05, "epoch": 3.001364256480218, "percentage": 60.11, "elapsed_time": "8:10:33", "remaining_time": "5:25:32"}
 {"current_steps": 551, "total_steps": 915, "loss": 0.86, "lr": 3.278691026210904e-05, "epoch": 3.0068212824010914, "percentage": 60.22, "elapsed_time": "8:11:40", "remaining_time": "5:24:48"}
+{"current_steps": 552, "total_steps": 915, "loss": 0.8687, "lr": 3.2636776472124144e-05, "epoch": 3.0122783083219646, "percentage": 60.33, "elapsed_time": "8:12:35", "remaining_time": "5:23:55"}
+{"current_steps": 553, "total_steps": 915, "loss": 0.8613, "lr": 3.248674997418652e-05, "epoch": 3.0177353342428375, "percentage": 60.44, "elapsed_time": "8:13:27", "remaining_time": "5:23:01"}
+{"current_steps": 554, "total_steps": 915, "loss": 0.8634, "lr": 3.233683295438356e-05, "epoch": 3.0231923601637107, "percentage": 60.55, "elapsed_time": "8:14:19", "remaining_time": "5:22:06"}
+{"current_steps": 555, "total_steps": 915, "loss": 0.8707, "lr": 3.2187027597207426e-05, "epoch": 3.028649386084584, "percentage": 60.66, "elapsed_time": "8:15:10", "remaining_time": "5:21:11"}
+{"current_steps": 556, "total_steps": 915, "loss": 0.8597, "lr": 3.2037336085523186e-05, "epoch": 3.034106412005457, "percentage": 60.77, "elapsed_time": "8:16:02", "remaining_time": "5:20:17"}
+{"current_steps": 557, "total_steps": 915, "loss": 0.8431, "lr": 3.188776060053702e-05, "epoch": 3.03956343792633, "percentage": 60.87, "elapsed_time": "8:16:54", "remaining_time": "5:19:22"}
+{"current_steps": 558, "total_steps": 915, "loss": 0.8809, "lr": 3.173830332176448e-05, "epoch": 3.0450204638472034, "percentage": 60.98, "elapsed_time": "8:17:46", "remaining_time": "5:18:28"}
+{"current_steps": 559, "total_steps": 915, "loss": 0.8648, "lr": 3.158896642699865e-05, "epoch": 3.050477489768076, "percentage": 61.09, "elapsed_time": "8:18:38", "remaining_time": "5:17:33"}
+{"current_steps": 560, "total_steps": 915, "loss": 0.8884, "lr": 3.1439752092278485e-05, "epoch": 3.0559345156889495, "percentage": 61.2, "elapsed_time": "8:19:30", "remaining_time": "5:16:39"}
+{"current_steps": 561, "total_steps": 915, "loss": 0.8681, "lr": 3.129066249185704e-05, "epoch": 3.0613915416098227, "percentage": 61.31, "elapsed_time": "8:20:22", "remaining_time": "5:15:44"}
+{"current_steps": 562, "total_steps": 915, "loss": 0.8716, "lr": 3.11416997981699e-05, "epoch": 3.0668485675306956, "percentage": 61.42, "elapsed_time": "8:21:14", "remaining_time": "5:14:49"}
+{"current_steps": 563, "total_steps": 915, "loss": 0.8685, "lr": 3.099286618180337e-05, "epoch": 3.072305593451569, "percentage": 61.53, "elapsed_time": "8:22:05", "remaining_time": "5:13:55"}
+{"current_steps": 564, "total_steps": 915, "loss": 0.8612, "lr": 3.084416381146297e-05, "epoch": 3.077762619372442, "percentage": 61.64, "elapsed_time": "8:22:58", "remaining_time": "5:13:01"}
+{"current_steps": 565, "total_steps": 915, "loss": 0.8906, "lr": 3.069559485394176e-05, "epoch": 3.083219645293315, "percentage": 61.75, "elapsed_time": "8:23:53", "remaining_time": "5:12:08"}
+{"current_steps": 566, "total_steps": 915, "loss": 0.8647, "lr": 3.054716147408883e-05, "epoch": 3.088676671214188, "percentage": 61.86, "elapsed_time": "8:25:00", "remaining_time": "5:11:23"}
+{"current_steps": 567, "total_steps": 915, "loss": 0.8734, "lr": 3.039886583477767e-05, "epoch": 3.0941336971350615, "percentage": 61.97, "elapsed_time": "8:25:55", "remaining_time": "5:10:31"}
+{"current_steps": 568, "total_steps": 915, "loss": 0.8571, "lr": 3.0250710096874778e-05, "epoch": 3.0995907230559343, "percentage": 62.08, "elapsed_time": "8:26:47", "remaining_time": "5:09:36"}
+{"current_steps": 569, "total_steps": 915, "loss": 0.8425, "lr": 3.0102696419208032e-05, "epoch": 3.1050477489768076, "percentage": 62.19, "elapsed_time": "8:27:39", "remaining_time": "5:08:42"}
+{"current_steps": 570, "total_steps": 915, "loss": 0.876, "lr": 2.995482695853535e-05, "epoch": 3.110504774897681, "percentage": 62.3, "elapsed_time": "8:28:31", "remaining_time": "5:07:47"}
+{"current_steps": 571, "total_steps": 915, "loss": 0.8728, "lr": 2.9807103869513184e-05, "epoch": 3.1159618008185537, "percentage": 62.4, "elapsed_time": "8:29:23", "remaining_time": "5:06:53"}
+{"current_steps": 572, "total_steps": 915, "loss": 0.8933, "lr": 2.965952930466517e-05, "epoch": 3.121418826739427, "percentage": 62.51, "elapsed_time": "8:30:15", "remaining_time": "5:05:58"}
+{"current_steps": 573, "total_steps": 915, "loss": 0.8888, "lr": 2.9512105414350722e-05, "epoch": 3.1268758526603, "percentage": 62.62, "elapsed_time": "8:31:07", "remaining_time": "5:05:04"}
+{"current_steps": 574, "total_steps": 915, "loss": 0.8791, "lr": 2.9364834346733784e-05, "epoch": 3.132332878581173, "percentage": 62.73, "elapsed_time": "8:31:59", "remaining_time": "5:04:09"}
+{"current_steps": 575, "total_steps": 915, "loss": 0.8767, "lr": 2.9217718247751407e-05, "epoch": 3.1377899045020463, "percentage": 62.84, "elapsed_time": "8:32:51", "remaining_time": "5:03:15"}
+{"current_steps": 576, "total_steps": 915, "loss": 0.879, "lr": 2.907075926108257e-05, "epoch": 3.1432469304229196, "percentage": 62.95, "elapsed_time": "8:33:43", "remaining_time": "5:02:20"}
+{"current_steps": 577, "total_steps": 915, "loss": 0.8824, "lr": 2.892395952811691e-05, "epoch": 3.148703956343793, "percentage": 63.06, "elapsed_time": "8:34:34", "remaining_time": "5:01:26"}
+{"current_steps": 578, "total_steps": 915, "loss": 0.8598, "lr": 2.8777321187923513e-05, "epoch": 3.1541609822646657, "percentage": 63.17, "elapsed_time": "8:35:26", "remaining_time": "5:00:31"}
+{"current_steps": 579, "total_steps": 915, "loss": 0.8637, "lr": 2.8630846377219756e-05, "epoch": 3.159618008185539, "percentage": 63.28, "elapsed_time": "8:36:19", "remaining_time": "4:59:37"}
+{"current_steps": 580, "total_steps": 915, "loss": 0.88, "lr": 2.848453723034021e-05, "epoch": 3.1650750341064118, "percentage": 63.39, "elapsed_time": "8:37:16", "remaining_time": "4:58:46"}
+{"current_steps": 581, "total_steps": 915, "loss": 0.8648, "lr": 2.8338395879205447e-05, "epoch": 3.170532060027285, "percentage": 63.5, "elapsed_time": "8:38:22", "remaining_time": "4:57:59"}
+{"current_steps": 582, "total_steps": 915, "loss": 0.8574, "lr": 2.819242445329108e-05, "epoch": 3.1759890859481583, "percentage": 63.61, "elapsed_time": "8:39:17", "remaining_time": "4:57:07"}
+{"current_steps": 583, "total_steps": 915, "loss": 0.8635, "lr": 2.8046625079596643e-05, "epoch": 3.1814461118690316, "percentage": 63.72, "elapsed_time": "8:40:09", "remaining_time": "4:56:12"}
+{"current_steps": 584, "total_steps": 915, "loss": 0.8559, "lr": 2.7900999882614663e-05, "epoch": 3.1869031377899044, "percentage": 63.83, "elapsed_time": "8:41:01", "remaining_time": "4:55:18"}
+{"current_steps": 585, "total_steps": 915, "loss": 0.8718, "lr": 2.7755550984299656e-05, "epoch": 3.1923601637107777, "percentage": 63.93, "elapsed_time": "8:41:53", "remaining_time": "4:54:23"}
+{"current_steps": 586, "total_steps": 915, "loss": 0.8677, "lr": 2.7610280504037236e-05, "epoch": 3.197817189631651, "percentage": 64.04, "elapsed_time": "8:42:45", "remaining_time": "4:53:29"}
+{"current_steps": 587, "total_steps": 915, "loss": 0.8715, "lr": 2.746519055861326e-05, "epoch": 3.203274215552524, "percentage": 64.15, "elapsed_time": "8:43:37", "remaining_time": "4:52:35"}
+{"current_steps": 588, "total_steps": 915, "loss": 0.8526, "lr": 2.732028326218292e-05, "epoch": 3.208731241473397, "percentage": 64.26, "elapsed_time": "8:44:28", "remaining_time": "4:51:40"}
+{"current_steps": 589, "total_steps": 915, "loss": 0.8542, "lr": 2.7175560726239972e-05, "epoch": 3.2141882673942703, "percentage": 64.37, "elapsed_time": "8:45:20", "remaining_time": "4:50:46"}
+{"current_steps": 590, "total_steps": 915, "loss": 0.8805, "lr": 2.7031025059585957e-05, "epoch": 3.219645293315143, "percentage": 64.48, "elapsed_time": "8:46:12", "remaining_time": "4:49:51"}
+{"current_steps": 591, "total_steps": 915, "loss": 0.879, "lr": 2.6886678368299478e-05, "epoch": 3.2251023192360164, "percentage": 64.59, "elapsed_time": "8:47:04", "remaining_time": "4:48:57"}
+{"current_steps": 592, "total_steps": 915, "loss": 0.8581, "lr": 2.674252275570552e-05, "epoch": 3.2305593451568897, "percentage": 64.7, "elapsed_time": "8:47:56", "remaining_time": "4:48:03"}
+{"current_steps": 593, "total_steps": 915, "loss": 0.8672, "lr": 2.659856032234482e-05, "epoch": 3.2360163710777625, "percentage": 64.81, "elapsed_time": "8:48:48", "remaining_time": "4:47:08"}
+{"current_steps": 594, "total_steps": 915, "loss": 0.8635, "lr": 2.6454793165943187e-05, "epoch": 3.241473396998636, "percentage": 64.92, "elapsed_time": "8:49:40", "remaining_time": "4:46:14"}
+{"current_steps": 595, "total_steps": 915, "loss": 0.8596, "lr": 2.631122338138103e-05, "epoch": 3.246930422919509, "percentage": 65.03, "elapsed_time": "8:50:38", "remaining_time": "4:45:23"}
+{"current_steps": 596, "total_steps": 915, "loss": 0.8886, "lr": 2.6167853060662743e-05, "epoch": 3.252387448840382, "percentage": 65.14, "elapsed_time": "8:51:44", "remaining_time": "4:44:36"}
+{"current_steps": 597, "total_steps": 915, "loss": 0.8508, "lr": 2.6024684292886288e-05, "epoch": 3.257844474761255, "percentage": 65.25, "elapsed_time": "8:52:39", "remaining_time": "4:43:43"}
+{"current_steps": 598, "total_steps": 915, "loss": 0.8578, "lr": 2.5881719164212714e-05, "epoch": 3.2633015006821284, "percentage": 65.36, "elapsed_time": "8:53:31", "remaining_time": "4:42:49"}
+{"current_steps": 599, "total_steps": 915, "loss": 0.8555, "lr": 2.5738959757835807e-05, "epoch": 3.2687585266030013, "percentage": 65.46, "elapsed_time": "8:54:22", "remaining_time": "4:41:54"}
+{"current_steps": 600, "total_steps": 915, "loss": 0.892, "lr": 2.559640815395167e-05, "epoch": 3.2742155525238745, "percentage": 65.57, "elapsed_time": "8:55:14", "remaining_time": "4:41:00"}
+{"current_steps": 601, "total_steps": 915, "loss": 0.8568, "lr": 2.5454066429728465e-05, "epoch": 3.279672578444748, "percentage": 65.68, "elapsed_time": "8:56:06", "remaining_time": "4:40:05"}
+{"current_steps": 602, "total_steps": 915, "loss": 0.8494, "lr": 2.531193665927612e-05, "epoch": 3.2851296043656206, "percentage": 65.79, "elapsed_time": "8:56:58", "remaining_time": "4:39:11"}
+{"current_steps": 603, "total_steps": 915, "loss": 0.8653, "lr": 2.5170020913616092e-05, "epoch": 3.290586630286494, "percentage": 65.9, "elapsed_time": "8:57:50", "remaining_time": "4:38:17"}
+{"current_steps": 604, "total_steps": 915, "loss": 0.8653, "lr": 2.5028321260651235e-05, "epoch": 3.296043656207367, "percentage": 66.01, "elapsed_time": "8:58:42", "remaining_time": "4:37:23"}
+{"current_steps": 605, "total_steps": 915, "loss": 0.8519, "lr": 2.4886839765135643e-05, "epoch": 3.30150068212824, "percentage": 66.12, "elapsed_time": "8:59:34", "remaining_time": "4:36:28"}
+{"current_steps": 606, "total_steps": 915, "loss": 0.8746, "lr": 2.4745578488644544e-05, "epoch": 3.3069577080491133, "percentage": 66.23, "elapsed_time": "9:00:26", "remaining_time": "4:35:34"}
+{"current_steps": 607, "total_steps": 915, "loss": 0.8553, "lr": 2.460453948954431e-05, "epoch": 3.3124147339699865, "percentage": 66.34, "elapsed_time": "9:01:18", "remaining_time": "4:34:40"}
+{"current_steps": 608, "total_steps": 915, "loss": 0.8804, "lr": 2.44637248229624e-05, "epoch": 3.3178717598908594, "percentage": 66.45, "elapsed_time": "9:02:10", "remaining_time": "4:33:45"}
+{"current_steps": 609, "total_steps": 915, "loss": 0.8563, "lr": 2.432313654075746e-05, "epoch": 3.3233287858117326, "percentage": 66.56, "elapsed_time": "9:03:02", "remaining_time": "4:32:51"}
+{"current_steps": 610, "total_steps": 915, "loss": 0.8845, "lr": 2.4182776691489407e-05, "epoch": 3.328785811732606, "percentage": 66.67, "elapsed_time": "9:04:01", "remaining_time": "4:32:00"}
+{"current_steps": 611, "total_steps": 915, "loss": 0.87, "lr": 2.4042647320389597e-05, "epoch": 3.3342428376534787, "percentage": 66.78, "elapsed_time": "9:05:07", "remaining_time": "4:31:13"}
+{"current_steps": 612, "total_steps": 915, "loss": 0.8768, "lr": 2.3902750469331003e-05, "epoch": 3.339699863574352, "percentage": 66.89, "elapsed_time": "9:06:01", "remaining_time": "4:30:20"}
+{"current_steps": 613, "total_steps": 915, "loss": 0.8673, "lr": 2.3763088176798473e-05, "epoch": 3.3451568894952253, "percentage": 66.99, "elapsed_time": "9:06:53", "remaining_time": "4:29:26"}
+{"current_steps": 614, "total_steps": 915, "loss": 0.8764, "lr": 2.3623662477859016e-05, "epoch": 3.350613915416098, "percentage": 67.1, "elapsed_time": "9:07:45", "remaining_time": "4:28:31"}
+{"current_steps": 615, "total_steps": 915, "loss": 0.8594, "lr": 2.348447540413215e-05, "epoch": 3.3560709413369714, "percentage": 67.21, "elapsed_time": "9:08:37", "remaining_time": "4:27:37"}
+{"current_steps": 616, "total_steps": 915, "loss": 0.887, "lr": 2.3345528983760306e-05, "epoch": 3.3615279672578446, "percentage": 67.32, "elapsed_time": "9:09:29", "remaining_time": "4:26:43"}
+{"current_steps": 617, "total_steps": 915, "loss": 0.8726, "lr": 2.3206825241379275e-05, "epoch": 3.3669849931787175, "percentage": 67.43, "elapsed_time": "9:10:21", "remaining_time": "4:25:48"}
+{"current_steps": 618, "total_steps": 915, "loss": 0.8585, "lr": 2.3068366198088725e-05, "epoch": 3.3724420190995907, "percentage": 67.54, "elapsed_time": "9:11:13", "remaining_time": "4:24:54"}
+{"current_steps": 619, "total_steps": 915, "loss": 0.8686, "lr": 2.293015387142271e-05, "epoch": 3.377899045020464, "percentage": 67.65, "elapsed_time": "9:12:05", "remaining_time": "4:24:00"}
+{"current_steps": 620, "total_steps": 915, "loss": 0.8524, "lr": 2.2792190275320268e-05, "epoch": 3.383356070941337, "percentage": 67.76, "elapsed_time": "9:12:57", "remaining_time": "4:23:05"}
+{"current_steps": 621, "total_steps": 915, "loss": 0.8757, "lr": 2.265447742009615e-05, "epoch": 3.38881309686221, "percentage": 67.87, "elapsed_time": "9:13:49", "remaining_time": "4:22:11"}
+{"current_steps": 622, "total_steps": 915, "loss": 0.8682, "lr": 2.251701731241143e-05, "epoch": 3.3942701227830834, "percentage": 67.98, "elapsed_time": "9:14:41", "remaining_time": "4:21:17"}
+{"current_steps": 623, "total_steps": 915, "loss": 0.8658, "lr": 2.237981195524432e-05, "epoch": 3.399727148703956, "percentage": 68.09, "elapsed_time": "9:15:33", "remaining_time": "4:20:23"}
+{"current_steps": 624, "total_steps": 915, "loss": 0.857, "lr": 2.2242863347861018e-05, "epoch": 3.4051841746248295, "percentage": 68.2, "elapsed_time": "9:16:25", "remaining_time": "4:19:29"}
+{"current_steps": 625, "total_steps": 915, "loss": 0.8751, "lr": 2.210617348578649e-05, "epoch": 3.4106412005457027, "percentage": 68.31, "elapsed_time": "9:17:24", "remaining_time": "4:18:38"}
+{"current_steps": 626, "total_steps": 915, "loss": 0.8721, "lr": 2.196974436077543e-05, "epoch": 3.4160982264665756, "percentage": 68.42, "elapsed_time": "9:18:28", "remaining_time": "4:17:49"}
+{"current_steps": 627, "total_steps": 915, "loss": 0.8837, "lr": 2.1833577960783266e-05, "epoch": 3.421555252387449, "percentage": 68.52, "elapsed_time": "9:19:23", "remaining_time": "4:16:56"}
+{"current_steps": 628, "total_steps": 915, "loss": 0.8578, "lr": 2.1697676269937178e-05, "epoch": 3.427012278308322, "percentage": 68.63, "elapsed_time": "9:20:15", "remaining_time": "4:16:02"}
+{"current_steps": 629, "total_steps": 915, "loss": 0.8572, "lr": 2.1562041268507157e-05, "epoch": 3.432469304229195, "percentage": 68.74, "elapsed_time": "9:21:06", "remaining_time": "4:15:08"}
+{"current_steps": 630, "total_steps": 915, "loss": 0.8574, "lr": 2.1426674932877156e-05, "epoch": 3.437926330150068, "percentage": 68.85, "elapsed_time": "9:21:58", "remaining_time": "4:14:13"}
+{"current_steps": 631, "total_steps": 915, "loss": 0.8732, "lr": 2.1291579235516363e-05, "epoch": 3.4433833560709415, "percentage": 68.96, "elapsed_time": "9:22:50", "remaining_time": "4:13:19"}
+{"current_steps": 632, "total_steps": 915, "loss": 0.8643, "lr": 2.115675614495036e-05, "epoch": 3.4488403819918143, "percentage": 69.07, "elapsed_time": "9:23:42", "remaining_time": "4:12:25"}
+{"current_steps": 633, "total_steps": 915, "loss": 0.8554, "lr": 2.1022207625732477e-05, "epoch": 3.4542974079126876, "percentage": 69.18, "elapsed_time": "9:24:34", "remaining_time": "4:11:31"}
+{"current_steps": 634, "total_steps": 915, "loss": 0.8538, "lr": 2.0887935638415203e-05, "epoch": 3.459754433833561, "percentage": 69.29, "elapsed_time": "9:25:26", "remaining_time": "4:10:36"}
+{"current_steps": 635, "total_steps": 915, "loss": 0.8899, "lr": 2.075394213952155e-05, "epoch": 3.4652114597544337, "percentage": 69.4, "elapsed_time": "9:26:18", "remaining_time": "4:09:42"}
+{"current_steps": 636, "total_steps": 915, "loss": 0.87, "lr": 2.0620229081516596e-05, "epoch": 3.470668485675307, "percentage": 69.51, "elapsed_time": "9:27:10", "remaining_time": "4:08:48"}
+{"current_steps": 637, "total_steps": 915, "loss": 0.8776, "lr": 2.0486798412779053e-05, "epoch": 3.47612551159618, "percentage": 69.62, "elapsed_time": "9:28:02", "remaining_time": "4:07:54"}
+{"current_steps": 638, "total_steps": 915, "loss": 0.8562, "lr": 2.035365207757279e-05, "epoch": 3.481582537517053, "percentage": 69.73, "elapsed_time": "9:28:54", "remaining_time": "4:06:59"}
+{"current_steps": 639, "total_steps": 915, "loss": 0.8689, "lr": 2.022079201601856e-05, "epoch": 3.4870395634379263, "percentage": 69.84, "elapsed_time": "9:29:46", "remaining_time": "4:06:05"}
+{"current_steps": 640, "total_steps": 915, "loss": 0.8664, "lr": 2.008822016406578e-05, "epoch": 3.4924965893587996, "percentage": 69.95, "elapsed_time": "9:30:46", "remaining_time": "4:05:15"}
+{"current_steps": 641, "total_steps": 915, "loss": 0.8514, "lr": 1.9955938453464198e-05, "epoch": 3.4979536152796724, "percentage": 70.05, "elapsed_time": "9:31:51", "remaining_time": "4:04:26"}
+{"current_steps": 642, "total_steps": 915, "loss": 0.864, "lr": 1.9823948811735834e-05, "epoch": 3.5034106412005457, "percentage": 70.16, "elapsed_time": "9:32:45", "remaining_time": "4:03:33"}
+{"current_steps": 643, "total_steps": 915, "loss": 0.8726, "lr": 1.9692253162146913e-05, "epoch": 3.508867667121419, "percentage": 70.27, "elapsed_time": "9:33:37", "remaining_time": "4:02:39"}
+{"current_steps": 644, "total_steps": 915, "loss": 0.8738, "lr": 1.9560853423679754e-05, "epoch": 3.5143246930422922, "percentage": 70.38, "elapsed_time": "9:34:29", "remaining_time": "4:01:44"}
+{"current_steps": 645, "total_steps": 915, "loss": 0.8616, "lr": 1.9429751511004858e-05, "epoch": 3.519781718963165, "percentage": 70.49, "elapsed_time": "9:35:21", "remaining_time": "4:00:50"}
+{"current_steps": 646, "total_steps": 915, "loss": 0.892, "lr": 1.9298949334453026e-05, "epoch": 3.5252387448840383, "percentage": 70.6, "elapsed_time": "9:36:13", "remaining_time": "3:59:56"}
+{"current_steps": 647, "total_steps": 915, "loss": 0.8786, "lr": 1.916844879998748e-05, "epoch": 3.530695770804911, "percentage": 70.71, "elapsed_time": "9:37:04", "remaining_time": "3:59:02"}
+{"current_steps": 648, "total_steps": 915, "loss": 0.8858, "lr": 1.9038251809176117e-05, "epoch": 3.5361527967257844, "percentage": 70.82, "elapsed_time": "9:37:56", "remaining_time": "3:58:08"}
+{"current_steps": 649, "total_steps": 915, "loss": 0.8743, "lr": 1.890836025916382e-05, "epoch": 3.5416098226466577, "percentage": 70.93, "elapsed_time": "9:38:48", "remaining_time": "3:57:14"}
+{"current_steps": 650, "total_steps": 915, "loss": 0.8656, "lr": 1.8778776042644763e-05, "epoch": 3.547066848567531, "percentage": 71.04, "elapsed_time": "9:39:40", "remaining_time": "3:56:19"}
+{"current_steps": 651, "total_steps": 915, "loss": 0.8715, "lr": 1.864950104783484e-05, "epoch": 3.552523874488404, "percentage": 71.15, "elapsed_time": "9:40:32", "remaining_time": "3:55:25"}
+{"current_steps": 652, "total_steps": 915, "loss": 0.8813, "lr": 1.852053715844424e-05, "epoch": 3.557980900409277, "percentage": 71.26, "elapsed_time": "9:41:24", "remaining_time": "3:54:31"}
+{"current_steps": 653, "total_steps": 915, "loss": 0.8781, "lr": 1.839188625364986e-05, "epoch": 3.56343792633015, "percentage": 71.37, "elapsed_time": "9:42:16", "remaining_time": "3:53:37"}
+{"current_steps": 654, "total_steps": 915, "loss": 0.8655, "lr": 1.8263550208068018e-05, "epoch": 3.568894952251023, "percentage": 71.48, "elapsed_time": "9:43:08", "remaining_time": "3:52:43"}
+{"current_steps": 655, "total_steps": 915, "loss": 0.852, "lr": 1.813553089172709e-05, "epoch": 3.5743519781718964, "percentage": 71.58, "elapsed_time": "9:44:09", "remaining_time": "3:51:52"}
+{"current_steps": 656, "total_steps": 915, "loss": 0.8493, "lr": 1.800783017004034e-05, "epoch": 3.5798090040927697, "percentage": 71.69, "elapsed_time": "9:45:14", "remaining_time": "3:51:03"}
+{"current_steps": 657, "total_steps": 915, "loss": 0.8434, "lr": 1.7880449903778608e-05, "epoch": 3.5852660300136425, "percentage": 71.8, "elapsed_time": "9:46:07", "remaining_time": "3:50:10"}
+{"current_steps": 658, "total_steps": 915, "loss": 0.875, "lr": 1.7753391949043318e-05, "epoch": 3.590723055934516, "percentage": 71.91, "elapsed_time": "9:46:59", "remaining_time": "3:49:16"}
+{"current_steps": 659, "total_steps": 915, "loss": 0.8791, "lr": 1.7626658157239358e-05, "epoch": 3.5961800818553886, "percentage": 72.02, "elapsed_time": "9:47:51", "remaining_time": "3:48:21"}
+{"current_steps": 660, "total_steps": 915, "loss": 0.8338, "lr": 1.7500250375048136e-05, "epoch": 3.601637107776262, "percentage": 72.13, "elapsed_time": "9:48:43", "remaining_time": "3:47:27"}
+{"current_steps": 661, "total_steps": 915, "loss": 0.8831, "lr": 1.7374170444400645e-05, "epoch": 3.607094133697135, "percentage": 72.24, "elapsed_time": "9:49:35", "remaining_time": "3:46:33"}
+{"current_steps": 662, "total_steps": 915, "loss": 0.8725, "lr": 1.7248420202450687e-05, "epoch": 3.6125511596180084, "percentage": 72.35, "elapsed_time": "9:50:27", "remaining_time": "3:45:39"}
+{"current_steps": 663, "total_steps": 915, "loss": 0.8669, "lr": 1.7123001481548017e-05, "epoch": 3.6180081855388813, "percentage": 72.46, "elapsed_time": "9:51:19", "remaining_time": "3:44:45"}
+{"current_steps": 664, "total_steps": 915, "loss": 0.8524, "lr": 1.6997916109211674e-05, "epoch": 3.6234652114597545, "percentage": 72.57, "elapsed_time": "9:52:11", "remaining_time": "3:43:51"}
+{"current_steps": 665, "total_steps": 915, "loss": 0.8776, "lr": 1.6873165908103413e-05, "epoch": 3.6289222373806274, "percentage": 72.68, "elapsed_time": "9:53:03", "remaining_time": "3:42:57"}
+{"current_steps": 666, "total_steps": 915, "loss": 0.8508, "lr": 1.6748752696001047e-05, "epoch": 3.6343792633015006, "percentage": 72.79, "elapsed_time": "9:53:55", "remaining_time": "3:42:03"}
+{"current_steps": 667, "total_steps": 915, "loss": 0.8681, "lr": 1.6624678285772006e-05, "epoch": 3.639836289222374, "percentage": 72.9, "elapsed_time": "9:54:47", "remaining_time": "3:41:09"}
+{"current_steps": 668, "total_steps": 915, "loss": 0.8768, "lr": 1.6500944485346972e-05, "epoch": 3.645293315143247, "percentage": 73.01, "elapsed_time": "9:55:39", "remaining_time": "3:40:14"}
+{"current_steps": 669, "total_steps": 915, "loss": 0.8642, "lr": 1.6377553097693435e-05, "epoch": 3.65075034106412, "percentage": 73.11, "elapsed_time": "9:56:31", "remaining_time": "3:39:21"}
+{"current_steps": 670, "total_steps": 915, "loss": 0.8606, "lr": 1.625450592078949e-05, "epoch": 3.6562073669849933, "percentage": 73.22, "elapsed_time": "9:57:31", "remaining_time": "3:38:29"}
+{"current_steps": 671, "total_steps": 915, "loss": 0.8402, "lr": 1.6131804747597616e-05, "epoch": 3.661664392905866, "percentage": 73.33, "elapsed_time": "9:58:35", "remaining_time": "3:37:40"}
+{"current_steps": 672, "total_steps": 915, "loss": 0.8757, "lr": 1.600945136603855e-05, "epoch": 3.6671214188267394, "percentage": 73.44, "elapsed_time": "9:59:29", "remaining_time": "3:36:46"}
+{"current_steps": 673, "total_steps": 915, "loss": 0.8768, "lr": 1.5887447558965256e-05, "epoch": 3.6725784447476126, "percentage": 73.55, "elapsed_time": "10:00:21", "remaining_time": "3:35:52"}
+{"current_steps": 674, "total_steps": 915, "loss": 0.8776, "lr": 1.5765795104136894e-05, "epoch": 3.678035470668486, "percentage": 73.66, "elapsed_time": "10:01:13", "remaining_time": "3:34:58"}
+{"current_steps": 675, "total_steps": 915, "loss": 0.8729, "lr": 1.5644495774192998e-05, "epoch": 3.6834924965893587, "percentage": 73.77, "elapsed_time": "10:02:05", "remaining_time": "3:34:04"}
+{"current_steps": 676, "total_steps": 915, "loss": 0.8789, "lr": 1.552355133662755e-05, "epoch": 3.688949522510232, "percentage": 73.88, "elapsed_time": "10:02:57", "remaining_time": "3:33:10"}
+{"current_steps": 677, "total_steps": 915, "loss": 0.8548, "lr": 1.540296355376332e-05, "epoch": 3.694406548431105, "percentage": 73.99, "elapsed_time": "10:03:49", "remaining_time": "3:32:16"}
+{"current_steps": 678, "total_steps": 915, "loss": 0.8503, "lr": 1.5282734182726113e-05, "epoch": 3.699863574351978, "percentage": 74.1, "elapsed_time": "10:04:41", "remaining_time": "3:31:22"}
+{"current_steps": 679, "total_steps": 915, "loss": 0.87, "lr": 1.516286497541919e-05, "epoch": 3.7053206002728514, "percentage": 74.21, "elapsed_time": "10:05:33", "remaining_time": "3:30:28"}
+{"current_steps": 680, "total_steps": 915, "loss": 0.8506, "lr": 1.5043357678497734e-05, "epoch": 3.7107776261937246, "percentage": 74.32, "elapsed_time": "10:06:25", "remaining_time": "3:29:34"}
+{"current_steps": 681, "total_steps": 915, "loss": 0.8746, "lr": 1.4924214033343454e-05, "epoch": 3.7162346521145975, "percentage": 74.43, "elapsed_time": "10:07:16", "remaining_time": "3:28:40"}
+{"current_steps": 682, "total_steps": 915, "loss": 0.8776, "lr": 1.4805435776039123e-05, "epoch": 3.7216916780354707, "percentage": 74.54, "elapsed_time": "10:08:08", "remaining_time": "3:27:46"}
+{"current_steps": 683, "total_steps": 915, "loss": 0.8732, "lr": 1.4687024637343337e-05, "epoch": 3.7271487039563436, "percentage": 74.64, "elapsed_time": "10:09:00", "remaining_time": "3:26:52"}
+{"current_steps": 684, "total_steps": 915, "loss": 0.8782, "lr": 1.4568982342665274e-05, "epoch": 3.732605729877217, "percentage": 74.75, "elapsed_time": "10:09:53", "remaining_time": "3:25:58"}
+{"current_steps": 685, "total_steps": 915, "loss": 0.8615, "lr": 1.4451310612039566e-05, "epoch": 3.73806275579809, "percentage": 74.86, "elapsed_time": "10:10:56", "remaining_time": "3:25:07"}
+{"current_steps": 686, "total_steps": 915, "loss": 0.8755, "lr": 1.4334011160101216e-05, "epoch": 3.7435197817189634, "percentage": 74.97, "elapsed_time": "10:11:59", "remaining_time": "3:24:17"}
+{"current_steps": 687, "total_steps": 915, "loss": 0.8513, "lr": 1.4217085696060667e-05, "epoch": 3.748976807639836, "percentage": 75.08, "elapsed_time": "10:12:53", "remaining_time": "3:23:24"}
+{"current_steps": 688, "total_steps": 915, "loss": 0.8649, "lr": 1.4100535923678815e-05, "epoch": 3.7544338335607095, "percentage": 75.19, "elapsed_time": "10:13:45", "remaining_time": "3:22:30"}
+{"current_steps": 689, "total_steps": 915, "loss": 0.8616, "lr": 1.3984363541242218e-05, "epoch": 3.7598908594815823, "percentage": 75.3, "elapsed_time": "10:14:36", "remaining_time": "3:21:36"}
+{"current_steps": 690, "total_steps": 915, "loss": 0.8729, "lr": 1.3868570241538412e-05, "epoch": 3.7653478854024556, "percentage": 75.41, "elapsed_time": "10:15:28", "remaining_time": "3:20:42"}
+{"current_steps": 691, "total_steps": 915, "loss": 0.8792, "lr": 1.375315771183114e-05, "epoch": 3.770804911323329, "percentage": 75.52, "elapsed_time": "10:16:20", "remaining_time": "3:19:47"}
+{"current_steps": 692, "total_steps": 915, "loss": 0.8514, "lr": 1.3638127633835808e-05, "epoch": 3.776261937244202, "percentage": 75.63, "elapsed_time": "10:17:12", "remaining_time": "3:18:53"}
+{"current_steps": 693, "total_steps": 915, "loss": 0.8569, "lr": 1.3523481683695034e-05, "epoch": 3.781718963165075, "percentage": 75.74, "elapsed_time": "10:18:04", "remaining_time": "3:17:59"}
+{"current_steps": 694, "total_steps": 915, "loss": 0.8642, "lr": 1.3409221531954146e-05, "epoch": 3.787175989085948, "percentage": 75.85, "elapsed_time": "10:18:56", "remaining_time": "3:17:05"}
+{"current_steps": 695, "total_steps": 915, "loss": 0.8617, "lr": 1.329534884353687e-05, "epoch": 3.792633015006821, "percentage": 75.96, "elapsed_time": "10:19:48", "remaining_time": "3:16:11"}
+{"current_steps": 696, "total_steps": 915, "loss": 0.8957, "lr": 1.3181865277721086e-05, "epoch": 3.7980900409276943, "percentage": 76.07, "elapsed_time": "10:20:40", "remaining_time": "3:15:17"}
+{"current_steps": 697, "total_steps": 915, "loss": 0.8534, "lr": 1.3068772488114619e-05, "epoch": 3.8035470668485676, "percentage": 76.17, "elapsed_time": "10:21:32", "remaining_time": "3:14:23"}
+{"current_steps": 698, "total_steps": 915, "loss": 0.8767, "lr": 1.2956072122631173e-05, "epoch": 3.809004092769441, "percentage": 76.28, "elapsed_time": "10:22:24", "remaining_time": "3:13:29"}
+{"current_steps": 699, "total_steps": 915, "loss": 0.8817, "lr": 1.2843765823466292e-05, "epoch": 3.8144611186903137, "percentage": 76.39, "elapsed_time": "10:23:17", "remaining_time": "3:12:36"}
+{"current_steps": 700, "total_steps": 915, "loss": 0.8577, "lr": 1.2731855227073475e-05, "epoch": 3.819918144611187, "percentage": 76.5, "elapsed_time": "10:24:20", "remaining_time": "3:11:45"}
+{"current_steps": 701, "total_steps": 915, "loss": 0.8767, "lr": 1.2620341964140264e-05, "epoch": 3.8253751705320598, "percentage": 76.61, "elapsed_time": "10:25:22", "remaining_time": "3:10:54"}
+{"current_steps": 702, "total_steps": 915, "loss": 0.8604, "lr": 1.2509227659564548e-05, "epoch": 3.830832196452933, "percentage": 76.72, "elapsed_time": "10:26:16", "remaining_time": "3:10:01"}
+{"current_steps": 703, "total_steps": 915, "loss": 0.88, "lr": 1.2398513932430847e-05, "epoch": 3.8362892223738063, "percentage": 76.83, "elapsed_time": "10:27:08", "remaining_time": "3:09:07"}
+{"current_steps": 704, "total_steps": 915, "loss": 0.8705, "lr": 1.2288202395986723e-05, "epoch": 3.8417462482946796, "percentage": 76.94, "elapsed_time": "10:27:59", "remaining_time": "3:08:13"}
+{"current_steps": 705, "total_steps": 915, "loss": 0.8938, "lr": 1.2178294657619274e-05, "epoch": 3.8472032742155524, "percentage": 77.05, "elapsed_time": "10:28:51", "remaining_time": "3:07:19"}
+{"current_steps": 706, "total_steps": 915, "loss": 0.8786, "lr": 1.2068792318831748e-05, "epoch": 3.8526603001364257, "percentage": 77.16, "elapsed_time": "10:29:43", "remaining_time": "3:06:25"}
+{"current_steps": 707, "total_steps": 915, "loss": 0.8731, "lr": 1.1959696975220156e-05, "epoch": 3.8581173260572985, "percentage": 77.27, "elapsed_time": "10:30:35", "remaining_time": "3:05:31"}
+{"current_steps": 708, "total_steps": 915, "loss": 0.8544, "lr": 1.185101021645004e-05, "epoch": 3.863574351978172, "percentage": 77.38, "elapsed_time": "10:31:27", "remaining_time": "3:04:37"}
+{"current_steps": 709, "total_steps": 915, "loss": 0.8738, "lr": 1.1742733626233322e-05, "epoch": 3.869031377899045, "percentage": 77.49, "elapsed_time": "10:32:19", "remaining_time": "3:03:43"}
+{"current_steps": 710, "total_steps": 915, "loss": 0.8655, "lr": 1.163486878230521e-05, "epoch": 3.8744884038199183, "percentage": 77.6, "elapsed_time": "10:33:11", "remaining_time": "3:02:49"}
+{"current_steps": 711, "total_steps": 915, "loss": 0.8625, "lr": 1.1527417256401217e-05, "epoch": 3.879945429740791, "percentage": 77.7, "elapsed_time": "10:34:03", "remaining_time": "3:01:55"}
+{"current_steps": 712, "total_steps": 915, "loss": 0.8691, "lr": 1.1420380614234277e-05, "epoch": 3.8854024556616644, "percentage": 77.81, "elapsed_time": "10:34:55", "remaining_time": "3:01:01"}
+{"current_steps": 713, "total_steps": 915, "loss": 0.8613, "lr": 1.1313760415471875e-05, "epoch": 3.8908594815825372, "percentage": 77.92, "elapsed_time": "10:35:47", "remaining_time": "3:00:07"}
+{"current_steps": 714, "total_steps": 915, "loss": 0.8811, "lr": 1.1207558213713398e-05, "epoch": 3.8963165075034105, "percentage": 78.03, "elapsed_time": "10:36:40", "remaining_time": "2:59:13"}
+{"current_steps": 715, "total_steps": 915, "loss": 0.8665, "lr": 1.1101775556467427e-05, "epoch": 3.901773533424284, "percentage": 78.14, "elapsed_time": "10:37:42", "remaining_time": "2:58:22"}
+{"current_steps": 716, "total_steps": 915, "loss": 0.8894, "lr": 1.0996413985129224e-05, "epoch": 3.907230559345157, "percentage": 78.25, "elapsed_time": "10:38:46", "remaining_time": "2:57:32"}
+{"current_steps": 717, "total_steps": 915, "loss": 0.8623, "lr": 1.0891475034958265e-05, "epoch": 3.91268758526603, "percentage": 78.36, "elapsed_time": "10:39:39", "remaining_time": "2:56:38"}
+{"current_steps": 718, "total_steps": 915, "loss": 0.8882, "lr": 1.0786960235055855e-05, "epoch": 3.918144611186903, "percentage": 78.47, "elapsed_time": "10:40:31", "remaining_time": "2:55:44"}
+{"current_steps": 719, "total_steps": 915, "loss": 0.857, "lr": 1.0682871108342896e-05, "epoch": 3.923601637107776, "percentage": 78.58, "elapsed_time": "10:41:23", "remaining_time": "2:54:50"}
+{"current_steps": 720, "total_steps": 915, "loss": 0.8576, "lr": 1.0579209171537626e-05, "epoch": 3.9290586630286493, "percentage": 78.69, "elapsed_time": "10:42:15", "remaining_time": "2:53:56"}
+{"current_steps": 721, "total_steps": 915, "loss": 0.8515, "lr": 1.0475975935133573e-05, "epoch": 3.9345156889495225, "percentage": 78.8, "elapsed_time": "10:43:07", "remaining_time": "2:53:02"}
+{"current_steps": 722, "total_steps": 915, "loss": 0.8692, "lr": 1.0373172903377507e-05, "epoch": 3.939972714870396, "percentage": 78.91, "elapsed_time": "10:43:59", "remaining_time": "2:52:08"}
+{"current_steps": 723, "total_steps": 915, "loss": 0.8788, "lr": 1.0270801574247553e-05, "epoch": 3.9454297407912686, "percentage": 79.02, "elapsed_time": "10:44:51", "remaining_time": "2:51:14"}
+{"current_steps": 724, "total_steps": 915, "loss": 0.8653, "lr": 1.0168863439431336e-05, "epoch": 3.950886766712142, "percentage": 79.13, "elapsed_time": "10:45:42", "remaining_time": "2:50:20"}
+{"current_steps": 725, "total_steps": 915, "loss": 0.8676, "lr": 1.0067359984304286e-05, "epoch": 3.956343792633015, "percentage": 79.23, "elapsed_time": "10:46:34", "remaining_time": "2:49:26"}
+{"current_steps": 726, "total_steps": 915, "loss": 0.853, "lr": 9.966292687907928e-06, "epoch": 3.961800818553888, "percentage": 79.34, "elapsed_time": "10:47:26", "remaining_time": "2:48:33"}
+{"current_steps": 727, "total_steps": 915, "loss": 0.8605, "lr": 9.865663022928413e-06, "epoch": 3.9672578444747613, "percentage": 79.45, "elapsed_time": "10:48:18", "remaining_time": "2:47:39"}
+{"current_steps": 728, "total_steps": 915, "loss": 0.8682, "lr": 9.76547245567498e-06, "epoch": 3.9727148703956345, "percentage": 79.56, "elapsed_time": "10:49:10", "remaining_time": "2:46:45"}
+{"current_steps": 729, "total_steps": 915, "loss": 0.8555, "lr": 9.66572244605864e-06, "epoch": 3.9781718963165074, "percentage": 79.67, "elapsed_time": "10:50:03", "remaining_time": "2:45:51"}
+{"current_steps": 730, "total_steps": 915, "loss": 0.8697, "lr": 9.566414447570861e-06, "epoch": 3.9836289222373806, "percentage": 79.78, "elapsed_time": "10:51:05", "remaining_time": "2:45:00"}
+{"current_steps": 731, "total_steps": 915, "loss": 0.8709, "lr": 9.467549907262476e-06, "epoch": 3.989085948158254, "percentage": 79.89, "elapsed_time": "10:52:08", "remaining_time": "2:44:08"}
+{"current_steps": 732, "total_steps": 915, "loss": 0.8664, "lr": 9.36913026572248e-06, "epoch": 3.9945429740791267, "percentage": 80.0, "elapsed_time": "10:53:00", "remaining_time": "2:43:15"}
+{"current_steps": 733, "total_steps": 915, "loss": 1.5027, "lr": 9.271156957057123e-06, "epoch": 4.0, "percentage": 80.11, "elapsed_time": "10:53:51", "remaining_time": "2:42:21"}
+{"current_steps": 734, "total_steps": 915, "loss": 0.85, "lr": 9.173631408868976e-06, "epoch": 4.005457025920873, "percentage": 80.22, "elapsed_time": "10:56:04", "remaining_time": "2:41:47"}
+{"current_steps": 735, "total_steps": 915, "loss": 0.8465, "lr": 9.076555042236142e-06, "epoch": 4.0109140518417465, "percentage": 80.33, "elapsed_time": "10:56:56", "remaining_time": "2:40:53"}