Training in progress, epoch 4

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e619f1a4e60b9331928839674e223ecd83f14108b5779d8ebda6dadf2d97bac7
+oid sha256:7fc74972dac6783584914d2f1681dd747856c105e0d71003b4a5fe6d6dbb57ca
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:42b7973cff55e3677a466192e7153bc9626981b7ba026266d1ab63dbb1520dbb
+oid sha256:f31cd94210f7450365fe110ff130da18c9fcd7ea126b3176a9965b62650a093c
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f6f4e9c61bae96088e626f94739e63811186fce9c445a65db81882ef4d528c14
+oid sha256:393d2f35d92a233a855c6b7831c3d3ab9cd0ff620c829c51a53f3099e1581dba
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c318eeafbd1c83d29e35abfb8c42c486a16a07b1cf61fb3a3556ab2daf5ba2fa
+oid sha256:301d10fa388c2590c8b01c742665634f5de1fe1be1467a891c7bef7ca81a47b5
 size 1089994880

trainer_log.jsonl

@@ -633,3 +633,149 @@
 {"current_steps": 633, "total_steps": 780, "loss": 0.364, "lr": 1.0434718646516917e-06, "epoch": 4.044728434504792, "percentage": 81.15, "elapsed_time": "3:43:02", "remaining_time": "0:51:47"}
 {"current_steps": 634, "total_steps": 780, "loss": 0.3535, "lr": 1.0298303680084448e-06, "epoch": 4.05111821086262, "percentage": 81.28, "elapsed_time": "3:43:26", "remaining_time": "0:51:27"}
 {"current_steps": 635, "total_steps": 780, "loss": 0.3182, "lr": 1.0162683835815706e-06, "epoch": 4.057507987220447, "percentage": 81.41, "elapsed_time": "3:43:44", "remaining_time": "0:51:05"}
+{"current_steps": 636, "total_steps": 780, "loss": 0.3218, "lr": 1.0027861829824953e-06, "epoch": 4.063897763578275, "percentage": 81.54, "elapsed_time": "3:44:06", "remaining_time": "0:50:44"}
+{"current_steps": 637, "total_steps": 780, "loss": 0.3546, "lr": 9.893840362247809e-07, "epoch": 4.0702875399361025, "percentage": 81.67, "elapsed_time": "3:44:24", "remaining_time": "0:50:22"}
+{"current_steps": 638, "total_steps": 780, "loss": 0.3304, "lr": 9.760622117187234e-07, "epoch": 4.07667731629393, "percentage": 81.79, "elapsed_time": "3:44:47", "remaining_time": "0:50:01"}
+{"current_steps": 639, "total_steps": 780, "loss": 0.3341, "lr": 9.628209762659658e-07, "epoch": 4.083067092651757, "percentage": 81.92, "elapsed_time": "3:45:11", "remaining_time": "0:49:41"}
+{"current_steps": 640, "total_steps": 780, "loss": 0.2994, "lr": 9.496605950541676e-07, "epoch": 4.0894568690095845, "percentage": 82.05, "elapsed_time": "3:45:29", "remaining_time": "0:49:19"}
+{"current_steps": 641, "total_steps": 780, "loss": 0.2974, "lr": 9.365813316516787e-07, "epoch": 4.095846645367412, "percentage": 82.18, "elapsed_time": "3:45:49", "remaining_time": "0:48:58"}
+{"current_steps": 642, "total_steps": 780, "loss": 0.3594, "lr": 9.235834480022788e-07, "epoch": 4.102236421725239, "percentage": 82.31, "elapsed_time": "3:46:11", "remaining_time": "0:48:37"}
+{"current_steps": 643, "total_steps": 780, "loss": 0.3063, "lr": 9.10667204419915e-07, "epoch": 4.108626198083067, "percentage": 82.44, "elapsed_time": "3:46:30", "remaining_time": "0:48:15"}
+{"current_steps": 644, "total_steps": 780, "loss": 0.3199, "lr": 8.978328595834984e-07, "epoch": 4.115015974440895, "percentage": 82.56, "elapsed_time": "3:46:51", "remaining_time": "0:47:54"}
+{"current_steps": 645, "total_steps": 780, "loss": 0.339, "lr": 8.850806705317183e-07, "epoch": 4.121405750798722, "percentage": 82.69, "elapsed_time": "3:47:15", "remaining_time": "0:47:33"}
+{"current_steps": 646, "total_steps": 780, "loss": 0.3199, "lr": 8.724108926579e-07, "epoch": 4.127795527156549, "percentage": 82.82, "elapsed_time": "3:47:35", "remaining_time": "0:47:12"}
+{"current_steps": 647, "total_steps": 780, "loss": 0.3131, "lr": 8.598237797048825e-07, "epoch": 4.134185303514377, "percentage": 82.95, "elapsed_time": "3:47:55", "remaining_time": "0:46:51"}
+{"current_steps": 648, "total_steps": 780, "loss": 0.3114, "lr": 8.473195837599419e-07, "epoch": 4.140575079872204, "percentage": 83.08, "elapsed_time": "3:48:15", "remaining_time": "0:46:29"}
+{"current_steps": 649, "total_steps": 780, "loss": 0.3211, "lr": 8.348985552497424e-07, "epoch": 4.146964856230032, "percentage": 83.21, "elapsed_time": "3:48:37", "remaining_time": "0:46:08"}
+{"current_steps": 650, "total_steps": 780, "loss": 0.343, "lr": 8.225609429353187e-07, "epoch": 4.15335463258786, "percentage": 83.33, "elapsed_time": "3:48:59", "remaining_time": "0:45:47"}
+{"current_steps": 651, "total_steps": 780, "loss": 0.3035, "lr": 8.103069939070945e-07, "epoch": 4.159744408945687, "percentage": 83.46, "elapsed_time": "3:49:17", "remaining_time": "0:45:26"}
+{"current_steps": 652, "total_steps": 780, "loss": 0.3549, "lr": 7.981369535799354e-07, "epoch": 4.166134185303514, "percentage": 83.59, "elapsed_time": "3:49:41", "remaining_time": "0:45:05"}
+{"current_steps": 653, "total_steps": 780, "loss": 0.3351, "lr": 7.860510656882342e-07, "epoch": 4.172523961661342, "percentage": 83.72, "elapsed_time": "3:49:59", "remaining_time": "0:44:43"}
+{"current_steps": 654, "total_steps": 780, "loss": 0.3191, "lr": 7.740495722810271e-07, "epoch": 4.178913738019169, "percentage": 83.85, "elapsed_time": "3:50:21", "remaining_time": "0:44:22"}
+{"current_steps": 655, "total_steps": 780, "loss": 0.3583, "lr": 7.621327137171447e-07, "epoch": 4.185303514376997, "percentage": 83.97, "elapsed_time": "3:50:41", "remaining_time": "0:44:01"}
+{"current_steps": 656, "total_steps": 780, "loss": 0.342, "lr": 7.50300728660407e-07, "epoch": 4.1916932907348246, "percentage": 84.1, "elapsed_time": "3:51:06", "remaining_time": "0:43:41"}
+{"current_steps": 657, "total_steps": 780, "loss": 0.3354, "lr": 7.385538540748327e-07, "epoch": 4.198083067092652, "percentage": 84.23, "elapsed_time": "3:51:25", "remaining_time": "0:43:19"}
+{"current_steps": 658, "total_steps": 780, "loss": 0.3146, "lr": 7.26892325219899e-07, "epoch": 4.204472843450479, "percentage": 84.36, "elapsed_time": "3:51:47", "remaining_time": "0:42:58"}
+{"current_steps": 659, "total_steps": 780, "loss": 0.3308, "lr": 7.153163756458287e-07, "epoch": 4.210862619808307, "percentage": 84.49, "elapsed_time": "3:52:11", "remaining_time": "0:42:37"}
+{"current_steps": 660, "total_steps": 780, "loss": 0.2986, "lr": 7.03826237188916e-07, "epoch": 4.217252396166134, "percentage": 84.62, "elapsed_time": "3:52:32", "remaining_time": "0:42:16"}
+{"current_steps": 661, "total_steps": 780, "loss": 0.2966, "lr": 6.924221399668785e-07, "epoch": 4.223642172523961, "percentage": 84.74, "elapsed_time": "3:52:52", "remaining_time": "0:41:55"}
+{"current_steps": 662, "total_steps": 780, "loss": 0.2944, "lr": 6.811043123742494e-07, "epoch": 4.2300319488817895, "percentage": 84.87, "elapsed_time": "3:53:13", "remaining_time": "0:41:34"}
+{"current_steps": 663, "total_steps": 780, "loss": 0.3312, "lr": 6.698729810778065e-07, "epoch": 4.236421725239617, "percentage": 85.0, "elapsed_time": "3:53:33", "remaining_time": "0:41:13"}
+{"current_steps": 664, "total_steps": 780, "loss": 0.3086, "lr": 6.587283710120324e-07, "epoch": 4.242811501597444, "percentage": 85.13, "elapsed_time": "3:53:56", "remaining_time": "0:40:52"}
+{"current_steps": 665, "total_steps": 780, "loss": 0.3085, "lr": 6.47670705374604e-07, "epoch": 4.2492012779552715, "percentage": 85.26, "elapsed_time": "3:54:16", "remaining_time": "0:40:30"}
+{"current_steps": 666, "total_steps": 780, "loss": 0.3126, "lr": 6.367002056219285e-07, "epoch": 4.255591054313099, "percentage": 85.38, "elapsed_time": "3:54:35", "remaining_time": "0:40:09"}
+{"current_steps": 667, "total_steps": 780, "loss": 0.2721, "lr": 6.258170914647077e-07, "epoch": 4.261980830670926, "percentage": 85.51, "elapsed_time": "3:54:50", "remaining_time": "0:39:47"}
+{"current_steps": 668, "total_steps": 780, "loss": 0.3138, "lr": 6.150215808635334e-07, "epoch": 4.268370607028754, "percentage": 85.64, "elapsed_time": "3:55:10", "remaining_time": "0:39:25"}
+{"current_steps": 669, "total_steps": 780, "loss": 0.2922, "lr": 6.043138900245277e-07, "epoch": 4.274760383386582, "percentage": 85.77, "elapsed_time": "3:55:29", "remaining_time": "0:39:04"}
+{"current_steps": 670, "total_steps": 780, "loss": 0.319, "lr": 5.936942333950063e-07, "epoch": 4.281150159744409, "percentage": 85.9, "elapsed_time": "3:55:51", "remaining_time": "0:38:43"}
+{"current_steps": 671, "total_steps": 780, "loss": 0.3196, "lr": 5.831628236591929e-07, "epoch": 4.287539936102236, "percentage": 86.03, "elapsed_time": "3:56:12", "remaining_time": "0:38:22"}
+{"current_steps": 672, "total_steps": 780, "loss": 0.3343, "lr": 5.727198717339511e-07, "epoch": 4.293929712460064, "percentage": 86.15, "elapsed_time": "3:56:32", "remaining_time": "0:38:00"}
+{"current_steps": 673, "total_steps": 780, "loss": 0.3519, "lr": 5.623655867645628e-07, "epoch": 4.300319488817891, "percentage": 86.28, "elapsed_time": "3:56:56", "remaining_time": "0:37:40"}
+{"current_steps": 674, "total_steps": 780, "loss": 0.3267, "lr": 5.521001761205441e-07, "epoch": 4.306709265175719, "percentage": 86.41, "elapsed_time": "3:57:17", "remaining_time": "0:37:19"}
+{"current_steps": 675, "total_steps": 780, "loss": 0.3255, "lr": 5.41923845391486e-07, "epoch": 4.313099041533547, "percentage": 86.54, "elapsed_time": "3:57:32", "remaining_time": "0:36:57"}
+{"current_steps": 676, "total_steps": 780, "loss": 0.3837, "lr": 5.318367983829393e-07, "epoch": 4.319488817891374, "percentage": 86.67, "elapsed_time": "3:57:56", "remaining_time": "0:36:36"}
+{"current_steps": 677, "total_steps": 780, "loss": 0.3477, "lr": 5.218392371123326e-07, "epoch": 4.325878594249201, "percentage": 86.79, "elapsed_time": "3:58:12", "remaining_time": "0:36:14"}
+{"current_steps": 678, "total_steps": 780, "loss": 0.3554, "lr": 5.119313618049309e-07, "epoch": 4.332268370607029, "percentage": 86.92, "elapsed_time": "3:58:34", "remaining_time": "0:35:53"}
+{"current_steps": 679, "total_steps": 780, "loss": 0.3007, "lr": 5.021133708898146e-07, "epoch": 4.338658146964856, "percentage": 87.05, "elapsed_time": "3:58:52", "remaining_time": "0:35:31"}
+{"current_steps": 680, "total_steps": 780, "loss": 0.316, "lr": 4.9238546099592e-07, "epoch": 4.345047923322683, "percentage": 87.18, "elapsed_time": "3:59:16", "remaining_time": "0:35:11"}
+{"current_steps": 681, "total_steps": 780, "loss": 0.3518, "lr": 4.827478269480895e-07, "epoch": 4.3514376996805115, "percentage": 87.31, "elapsed_time": "3:59:38", "remaining_time": "0:34:50"}
+{"current_steps": 682, "total_steps": 780, "loss": 0.3401, "lr": 4.732006617631729e-07, "epoch": 4.357827476038339, "percentage": 87.44, "elapsed_time": "4:00:01", "remaining_time": "0:34:29"}
+{"current_steps": 683, "total_steps": 780, "loss": 0.3218, "lr": 4.637441566461681e-07, "epoch": 4.364217252396166, "percentage": 87.56, "elapsed_time": "4:00:24", "remaining_time": "0:34:08"}
+{"current_steps": 684, "total_steps": 780, "loss": 0.3241, "lr": 4.54378500986381e-07, "epoch": 4.3706070287539935, "percentage": 87.69, "elapsed_time": "4:00:41", "remaining_time": "0:33:46"}
+{"current_steps": 685, "total_steps": 780, "loss": 0.2896, "lr": 4.451038823536441e-07, "epoch": 4.376996805111821, "percentage": 87.82, "elapsed_time": "4:01:02", "remaining_time": "0:33:25"}
+{"current_steps": 686, "total_steps": 780, "loss": 0.3804, "lr": 4.35920486494546e-07, "epoch": 4.383386581469648, "percentage": 87.95, "elapsed_time": "4:01:23", "remaining_time": "0:33:04"}
+{"current_steps": 687, "total_steps": 780, "loss": 0.3811, "lr": 4.268284973287273e-07, "epoch": 4.389776357827476, "percentage": 88.08, "elapsed_time": "4:01:47", "remaining_time": "0:32:43"}
+{"current_steps": 688, "total_steps": 780, "loss": 0.2939, "lr": 4.1782809694518533e-07, "epoch": 4.396166134185304, "percentage": 88.21, "elapsed_time": "4:02:10", "remaining_time": "0:32:23"}
+{"current_steps": 689, "total_steps": 780, "loss": 0.2969, "lr": 4.089194655986306e-07, "epoch": 4.402555910543131, "percentage": 88.33, "elapsed_time": "4:02:29", "remaining_time": "0:32:01"}
+{"current_steps": 690, "total_steps": 780, "loss": 0.3453, "lr": 4.001027817058789e-07, "epoch": 4.4089456869009584, "percentage": 88.46, "elapsed_time": "4:02:51", "remaining_time": "0:31:40"}
+{"current_steps": 691, "total_steps": 780, "loss": 0.3329, "lr": 3.9137822184227845e-07, "epoch": 4.415335463258786, "percentage": 88.59, "elapsed_time": "4:03:11", "remaining_time": "0:31:19"}
+{"current_steps": 692, "total_steps": 780, "loss": 0.3237, "lr": 3.8274596073816784e-07, "epoch": 4.421725239616613, "percentage": 88.72, "elapsed_time": "4:03:34", "remaining_time": "0:30:58"}
+{"current_steps": 693, "total_steps": 780, "loss": 0.3189, "lr": 3.742061712753825e-07, "epoch": 4.428115015974441, "percentage": 88.85, "elapsed_time": "4:03:53", "remaining_time": "0:30:37"}
+{"current_steps": 694, "total_steps": 780, "loss": 0.3359, "lr": 3.657590244837911e-07, "epoch": 4.434504792332269, "percentage": 88.97, "elapsed_time": "4:04:16", "remaining_time": "0:30:16"}
+{"current_steps": 695, "total_steps": 780, "loss": 0.3198, "lr": 3.5740468953786854e-07, "epoch": 4.440894568690096, "percentage": 89.1, "elapsed_time": "4:04:35", "remaining_time": "0:29:54"}
+{"current_steps": 696, "total_steps": 780, "loss": 0.3163, "lr": 3.49143333753309e-07, "epoch": 4.447284345047923, "percentage": 89.23, "elapsed_time": "4:04:53", "remaining_time": "0:29:33"}
+{"current_steps": 697, "total_steps": 780, "loss": 0.2897, "lr": 3.4097512258367385e-07, "epoch": 4.453674121405751, "percentage": 89.36, "elapsed_time": "4:05:17", "remaining_time": "0:29:12"}
+{"current_steps": 698, "total_steps": 780, "loss": 0.3399, "lr": 3.3290021961708163e-07, "epoch": 4.460063897763578, "percentage": 89.49, "elapsed_time": "4:05:40", "remaining_time": "0:28:51"}
+{"current_steps": 699, "total_steps": 780, "loss": 0.3296, "lr": 3.2491878657292643e-07, "epoch": 4.466453674121405, "percentage": 89.62, "elapsed_time": "4:06:03", "remaining_time": "0:28:30"}
+{"current_steps": 700, "total_steps": 780, "loss": 0.3343, "lr": 3.1703098329864237e-07, "epoch": 4.472843450479234, "percentage": 89.74, "elapsed_time": "4:06:28", "remaining_time": "0:28:10"}
+{"current_steps": 701, "total_steps": 780, "loss": 0.3468, "lr": 3.0923696776650414e-07, "epoch": 4.479233226837061, "percentage": 89.87, "elapsed_time": "4:06:49", "remaining_time": "0:27:49"}
+{"current_steps": 702, "total_steps": 780, "loss": 0.3345, "lr": 3.015368960704584e-07, "epoch": 4.485623003194888, "percentage": 90.0, "elapsed_time": "4:07:12", "remaining_time": "0:27:28"}
+{"current_steps": 703, "total_steps": 780, "loss": 0.3025, "lr": 2.9393092242300026e-07, "epoch": 4.492012779552716, "percentage": 90.13, "elapsed_time": "4:07:30", "remaining_time": "0:27:06"}
+{"current_steps": 704, "total_steps": 780, "loss": 0.3324, "lr": 2.864191991520848e-07, "epoch": 4.498402555910543, "percentage": 90.26, "elapsed_time": "4:07:52", "remaining_time": "0:26:45"}
+{"current_steps": 705, "total_steps": 780, "loss": 0.3147, "lr": 2.790018766980773e-07, "epoch": 4.50479233226837, "percentage": 90.38, "elapsed_time": "4:08:06", "remaining_time": "0:26:23"}
+{"current_steps": 706, "total_steps": 780, "loss": 0.3001, "lr": 2.71679103610738e-07, "epoch": 4.511182108626198, "percentage": 90.51, "elapsed_time": "4:08:27", "remaining_time": "0:26:02"}
+{"current_steps": 707, "total_steps": 780, "loss": 0.3057, "lr": 2.64451026546248e-07, "epoch": 4.517571884984026, "percentage": 90.64, "elapsed_time": "4:08:48", "remaining_time": "0:25:41"}
+{"current_steps": 708, "total_steps": 780, "loss": 0.2884, "lr": 2.573177902642726e-07, "epoch": 4.523961661341853, "percentage": 90.77, "elapsed_time": "4:09:10", "remaining_time": "0:25:20"}
+{"current_steps": 709, "total_steps": 780, "loss": 0.3257, "lr": 2.502795376250622e-07, "epoch": 4.5303514376996805, "percentage": 90.9, "elapsed_time": "4:09:32", "remaining_time": "0:24:59"}
+{"current_steps": 710, "total_steps": 780, "loss": 0.3223, "lr": 2.4333640958659144e-07, "epoch": 4.536741214057508, "percentage": 91.03, "elapsed_time": "4:09:50", "remaining_time": "0:24:37"}
+{"current_steps": 711, "total_steps": 780, "loss": 0.3332, "lr": 2.3648854520173237e-07, "epoch": 4.543130990415335, "percentage": 91.15, "elapsed_time": "4:10:06", "remaining_time": "0:24:16"}
+{"current_steps": 712, "total_steps": 780, "loss": 0.3146, "lr": 2.2973608161547755e-07, "epoch": 4.549520766773163, "percentage": 91.28, "elapsed_time": "4:10:26", "remaining_time": "0:23:55"}
+{"current_steps": 713, "total_steps": 780, "loss": 0.2865, "lr": 2.2307915406218517e-07, "epoch": 4.555910543130991, "percentage": 91.41, "elapsed_time": "4:10:46", "remaining_time": "0:23:33"}
+{"current_steps": 714, "total_steps": 780, "loss": 0.3136, "lr": 2.1651789586287442e-07, "epoch": 4.562300319488818, "percentage": 91.54, "elapsed_time": "4:11:04", "remaining_time": "0:23:12"}
+{"current_steps": 715, "total_steps": 780, "loss": 0.3043, "lr": 2.1005243842255552e-07, "epoch": 4.568690095846645, "percentage": 91.67, "elapsed_time": "4:11:28", "remaining_time": "0:22:51"}
+{"current_steps": 716, "total_steps": 780, "loss": 0.3074, "lr": 2.0368291122759898e-07, "epoch": 4.575079872204473, "percentage": 91.79, "elapsed_time": "4:11:46", "remaining_time": "0:22:30"}
+{"current_steps": 717, "total_steps": 780, "loss": 0.3133, "lr": 1.9740944184313882e-07, "epoch": 4.5814696485623, "percentage": 91.92, "elapsed_time": "4:12:10", "remaining_time": "0:22:09"}
+{"current_steps": 718, "total_steps": 780, "loss": 0.2996, "lr": 1.9123215591052014e-07, "epoch": 4.587859424920127, "percentage": 92.05, "elapsed_time": "4:12:30", "remaining_time": "0:21:48"}
+{"current_steps": 719, "total_steps": 780, "loss": 0.3463, "lr": 1.8515117714478447e-07, "epoch": 4.594249201277956, "percentage": 92.18, "elapsed_time": "4:12:53", "remaining_time": "0:21:27"}
+{"current_steps": 720, "total_steps": 780, "loss": 0.2938, "lr": 1.7916662733218848e-07, "epoch": 4.600638977635783, "percentage": 92.31, "elapsed_time": "4:13:14", "remaining_time": "0:21:06"}
+{"current_steps": 721, "total_steps": 780, "loss": 0.3073, "lr": 1.7327862632776638e-07, "epoch": 4.60702875399361, "percentage": 92.44, "elapsed_time": "4:13:27", "remaining_time": "0:20:44"}
+{"current_steps": 722, "total_steps": 780, "loss": 0.3487, "lr": 1.6748729205293024e-07, "epoch": 4.613418530351438, "percentage": 92.56, "elapsed_time": "4:13:47", "remaining_time": "0:20:23"}
+{"current_steps": 723, "total_steps": 780, "loss": 0.3433, "lr": 1.6179274049310966e-07, "epoch": 4.619808306709265, "percentage": 92.69, "elapsed_time": "4:14:07", "remaining_time": "0:20:02"}
+{"current_steps": 724, "total_steps": 780, "loss": 0.32, "lr": 1.5619508569542363e-07, "epoch": 4.626198083067092, "percentage": 92.82, "elapsed_time": "4:14:26", "remaining_time": "0:19:40"}
+{"current_steps": 725, "total_steps": 780, "loss": 0.3055, "lr": 1.5069443976640287e-07, "epoch": 4.63258785942492, "percentage": 92.95, "elapsed_time": "4:14:47", "remaining_time": "0:19:19"}
+{"current_steps": 726, "total_steps": 780, "loss": 0.3424, "lr": 1.4529091286973994e-07, "epoch": 4.638977635782748, "percentage": 93.08, "elapsed_time": "4:15:11", "remaining_time": "0:18:58"}
+{"current_steps": 727, "total_steps": 780, "loss": 0.3311, "lr": 1.3998461322408563e-07, "epoch": 4.645367412140575, "percentage": 93.21, "elapsed_time": "4:15:33", "remaining_time": "0:18:37"}
+{"current_steps": 728, "total_steps": 780, "loss": 0.324, "lr": 1.3477564710088097e-07, "epoch": 4.651757188498403, "percentage": 93.33, "elapsed_time": "4:15:57", "remaining_time": "0:18:16"}
+{"current_steps": 729, "total_steps": 780, "loss": 0.3216, "lr": 1.2966411882222695e-07, "epoch": 4.65814696485623, "percentage": 93.46, "elapsed_time": "4:16:16", "remaining_time": "0:17:55"}
+{"current_steps": 730, "total_steps": 780, "loss": 0.3529, "lr": 1.2465013075879884e-07, "epoch": 4.664536741214057, "percentage": 93.59, "elapsed_time": "4:16:37", "remaining_time": "0:17:34"}
+{"current_steps": 731, "total_steps": 780, "loss": 0.3343, "lr": 1.1973378332779229e-07, "epoch": 4.6709265175718855, "percentage": 93.72, "elapsed_time": "4:16:55", "remaining_time": "0:17:13"}
+{"current_steps": 732, "total_steps": 780, "loss": 0.3211, "lr": 1.1491517499091498e-07, "epoch": 4.677316293929713, "percentage": 93.85, "elapsed_time": "4:17:15", "remaining_time": "0:16:52"}
+{"current_steps": 733, "total_steps": 780, "loss": 0.3011, "lr": 1.1019440225241317e-07, "epoch": 4.68370607028754, "percentage": 93.97, "elapsed_time": "4:17:33", "remaining_time": "0:16:30"}
+{"current_steps": 734, "total_steps": 780, "loss": 0.3099, "lr": 1.055715596571405e-07, "epoch": 4.6900958466453675, "percentage": 94.1, "elapsed_time": "4:17:55", "remaining_time": "0:16:09"}
+{"current_steps": 735, "total_steps": 780, "loss": 0.2832, "lr": 1.0104673978866164e-07, "epoch": 4.696485623003195, "percentage": 94.23, "elapsed_time": "4:18:12", "remaining_time": "0:15:48"}
+{"current_steps": 736, "total_steps": 780, "loss": 0.3012, "lr": 9.662003326740166e-08, "epoch": 4.702875399361022, "percentage": 94.36, "elapsed_time": "4:18:36", "remaining_time": "0:15:27"}
+{"current_steps": 737, "total_steps": 780, "loss": 0.3747, "lr": 9.22915287488274e-08, "epoch": 4.7092651757188495, "percentage": 94.49, "elapsed_time": "4:18:58", "remaining_time": "0:15:06"}
+{"current_steps": 738, "total_steps": 780, "loss": 0.3206, "lr": 8.80613129216762e-08, "epoch": 4.715654952076678, "percentage": 94.62, "elapsed_time": "4:19:19", "remaining_time": "0:14:45"}
+{"current_steps": 739, "total_steps": 780, "loss": 0.3196, "lr": 8.392947050621603e-08, "epoch": 4.722044728434505, "percentage": 94.74, "elapsed_time": "4:19:43", "remaining_time": "0:14:24"}
+{"current_steps": 740, "total_steps": 780, "loss": 0.3322, "lr": 7.989608425254924e-08, "epoch": 4.728434504792332, "percentage": 94.87, "elapsed_time": "4:20:04", "remaining_time": "0:14:03"}
+{"current_steps": 741, "total_steps": 780, "loss": 0.335, "lr": 7.59612349389599e-08, "epoch": 4.73482428115016, "percentage": 95.0, "elapsed_time": "4:20:23", "remaining_time": "0:13:42"}
+{"current_steps": 742, "total_steps": 780, "loss": 0.3295, "lr": 7.212500137028789e-08, "epoch": 4.741214057507987, "percentage": 95.13, "elapsed_time": "4:20:45", "remaining_time": "0:13:21"}
+{"current_steps": 743, "total_steps": 780, "loss": 0.3209, "lr": 6.838746037635735e-08, "epoch": 4.747603833865814, "percentage": 95.26, "elapsed_time": "4:21:08", "remaining_time": "0:13:00"}
+{"current_steps": 744, "total_steps": 780, "loss": 0.292, "lr": 6.474868681043578e-08, "epoch": 4.753993610223642, "percentage": 95.38, "elapsed_time": "4:21:29", "remaining_time": "0:12:39"}
+{"current_steps": 745, "total_steps": 780, "loss": 0.3502, "lr": 6.120875354773459e-08, "epoch": 4.76038338658147, "percentage": 95.51, "elapsed_time": "4:21:52", "remaining_time": "0:12:18"}
+{"current_steps": 746, "total_steps": 780, "loss": 0.2825, "lr": 5.776773148394976e-08, "epoch": 4.766773162939297, "percentage": 95.64, "elapsed_time": "4:22:11", "remaining_time": "0:11:56"}
+{"current_steps": 747, "total_steps": 780, "loss": 0.3168, "lr": 5.4425689533841864e-08, "epoch": 4.773162939297125, "percentage": 95.77, "elapsed_time": "4:22:29", "remaining_time": "0:11:35"}
+{"current_steps": 748, "total_steps": 780, "loss": 0.3564, "lr": 5.1182694629857145e-08, "epoch": 4.779552715654952, "percentage": 95.9, "elapsed_time": "4:22:48", "remaining_time": "0:11:14"}
+{"current_steps": 749, "total_steps": 780, "loss": 0.3069, "lr": 4.803881172078473e-08, "epoch": 4.785942492012779, "percentage": 96.03, "elapsed_time": "4:23:06", "remaining_time": "0:10:53"}
+{"current_steps": 750, "total_steps": 780, "loss": 0.3187, "lr": 4.499410377045765e-08, "epoch": 4.792332268370607, "percentage": 96.15, "elapsed_time": "4:23:27", "remaining_time": "0:10:32"}
+{"current_steps": 751, "total_steps": 780, "loss": 0.3549, "lr": 4.2048631756492206e-08, "epoch": 4.798722044728435, "percentage": 96.28, "elapsed_time": "4:23:48", "remaining_time": "0:10:11"}
+{"current_steps": 752, "total_steps": 780, "loss": 0.2931, "lr": 3.9202454669063915e-08, "epoch": 4.805111821086262, "percentage": 96.41, "elapsed_time": "4:24:08", "remaining_time": "0:09:50"}
+{"current_steps": 753, "total_steps": 780, "loss": 0.3078, "lr": 3.645562950973014e-08, "epoch": 4.81150159744409, "percentage": 96.54, "elapsed_time": "4:24:28", "remaining_time": "0:09:29"}
+{"current_steps": 754, "total_steps": 780, "loss": 0.3127, "lr": 3.3808211290284886e-08, "epoch": 4.817891373801917, "percentage": 96.67, "elapsed_time": "4:24:47", "remaining_time": "0:09:07"}
+{"current_steps": 755, "total_steps": 780, "loss": 0.317, "lr": 3.1260253031660247e-08, "epoch": 4.824281150159744, "percentage": 96.79, "elapsed_time": "4:25:10", "remaining_time": "0:08:46"}
+{"current_steps": 756, "total_steps": 780, "loss": 0.3115, "lr": 2.8811805762860578e-08, "epoch": 4.830670926517572, "percentage": 96.92, "elapsed_time": "4:25:32", "remaining_time": "0:08:25"}
+{"current_steps": 757, "total_steps": 780, "loss": 0.3142, "lr": 2.6462918519944425e-08, "epoch": 4.8370607028754, "percentage": 97.05, "elapsed_time": "4:25:51", "remaining_time": "0:08:04"}
+{"current_steps": 758, "total_steps": 780, "loss": 0.3321, "lr": 2.4213638345040868e-08, "epoch": 4.843450479233227, "percentage": 97.18, "elapsed_time": "4:26:14", "remaining_time": "0:07:43"}
+{"current_steps": 759, "total_steps": 780, "loss": 0.2945, "lr": 2.206401028540639e-08, "epoch": 4.8498402555910545, "percentage": 97.31, "elapsed_time": "4:26:37", "remaining_time": "0:07:22"}
+{"current_steps": 760, "total_steps": 780, "loss": 0.3255, "lr": 2.0014077392525035e-08, "epoch": 4.856230031948882, "percentage": 97.44, "elapsed_time": "4:26:59", "remaining_time": "0:07:01"}
+{"current_steps": 761, "total_steps": 780, "loss": 0.3088, "lr": 1.8063880721242986e-08, "epoch": 4.862619808306709, "percentage": 97.56, "elapsed_time": "4:27:20", "remaining_time": "0:06:40"}
+{"current_steps": 762, "total_steps": 780, "loss": 0.2997, "lr": 1.6213459328950355e-08, "epoch": 4.8690095846645365, "percentage": 97.69, "elapsed_time": "4:27:42", "remaining_time": "0:06:19"}
+{"current_steps": 763, "total_steps": 780, "loss": 0.3173, "lr": 1.4462850274794548e-08, "epoch": 4.875399361022364, "percentage": 97.82, "elapsed_time": "4:28:04", "remaining_time": "0:05:58"}
+{"current_steps": 764, "total_steps": 780, "loss": 0.3333, "lr": 1.2812088618942009e-08, "epoch": 4.881789137380192, "percentage": 97.95, "elapsed_time": "4:28:28", "remaining_time": "0:05:37"}
+{"current_steps": 765, "total_steps": 780, "loss": 0.2929, "lr": 1.1261207421874309e-08, "epoch": 4.888178913738019, "percentage": 98.08, "elapsed_time": "4:28:49", "remaining_time": "0:05:16"}
+{"current_steps": 766, "total_steps": 780, "loss": 0.3297, "lr": 9.810237743724805e-09, "epoch": 4.894568690095847, "percentage": 98.21, "elapsed_time": "4:29:12", "remaining_time": "0:04:55"}
+{"current_steps": 767, "total_steps": 780, "loss": 0.339, "lr": 8.459208643659122e-09, "epoch": 4.900958466453674, "percentage": 98.33, "elapsed_time": "4:29:33", "remaining_time": "0:04:34"}
+{"current_steps": 768, "total_steps": 780, "loss": 0.3092, "lr": 7.2081471792911914e-09, "epoch": 4.907348242811501, "percentage": 98.46, "elapsed_time": "4:29:53", "remaining_time": "0:04:13"}
+{"current_steps": 769, "total_steps": 780, "loss": 0.3148, "lr": 6.057078406142003e-09, "epoch": 4.913738019169329, "percentage": 98.59, "elapsed_time": "4:30:15", "remaining_time": "0:03:51"}
+{"current_steps": 770, "total_steps": 780, "loss": 0.3567, "lr": 5.006025377138901e-09, "epoch": 4.920127795527156, "percentage": 98.72, "elapsed_time": "4:30:35", "remaining_time": "0:03:30"}
+{"current_steps": 771, "total_steps": 780, "loss": 0.2882, "lr": 4.055009142152066e-09, "epoch": 4.926517571884984, "percentage": 98.85, "elapsed_time": "4:30:51", "remaining_time": "0:03:09"}
+{"current_steps": 772, "total_steps": 780, "loss": 0.2884, "lr": 3.204048747573185e-09, "epoch": 4.932907348242812, "percentage": 98.97, "elapsed_time": "4:31:11", "remaining_time": "0:02:48"}
+{"current_steps": 773, "total_steps": 780, "loss": 0.3406, "lr": 2.4531612359363077e-09, "epoch": 4.939297124600639, "percentage": 99.1, "elapsed_time": "4:31:31", "remaining_time": "0:02:27"}
+{"current_steps": 774, "total_steps": 780, "loss": 0.3102, "lr": 1.8023616455731253e-09, "epoch": 4.945686900958466, "percentage": 99.23, "elapsed_time": "4:31:50", "remaining_time": "0:02:06"}
+{"current_steps": 775, "total_steps": 780, "loss": 0.3132, "lr": 1.2516630103137638e-09, "epoch": 4.952076677316294, "percentage": 99.36, "elapsed_time": "4:32:12", "remaining_time": "0:01:45"}
+{"current_steps": 776, "total_steps": 780, "loss": 0.3271, "lr": 8.010763592264381e-10, "epoch": 4.958466453674122, "percentage": 99.49, "elapsed_time": "4:32:33", "remaining_time": "0:01:24"}
+{"current_steps": 777, "total_steps": 780, "loss": 0.3287, "lr": 4.506107163948503e-10, "epoch": 4.964856230031949, "percentage": 99.62, "elapsed_time": "4:32:56", "remaining_time": "0:01:03"}
+{"current_steps": 778, "total_steps": 780, "loss": 0.3421, "lr": 2.0027310073833516e-10, "epoch": 4.9712460063897765, "percentage": 99.74, "elapsed_time": "4:33:17", "remaining_time": "0:00:42"}
+{"current_steps": 779, "total_steps": 780, "loss": 0.3439, "lr": 5.0068525870305974e-11, "epoch": 4.977635782747604, "percentage": 99.87, "elapsed_time": "4:33:35", "remaining_time": "0:00:21"}
+{"current_steps": 780, "total_steps": 780, "loss": 0.3082, "lr": 0.0, "epoch": 4.984025559105431, "percentage": 100.0, "elapsed_time": "4:33:58", "remaining_time": "0:00:00"}
+{"current_steps": 780, "total_steps": 780, "epoch": 4.984025559105431, "percentage": 100.0, "elapsed_time": "4:35:32", "remaining_time": "0:00:00"}