Training in progress, step 2500, checkpoint

checkpoint-2500/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a496b9e115dce09bc93e88a56d6a6d0e3f506e889c71b14136c5be5c2c4d7ba
+oid sha256:a4afbab03310aa655413f12f16bdda12fba4231c9b05e35602924db70f2f3080
 size 577789320

checkpoint-2500/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:be65299123489dd13c0da371c33ed033a03d520b1cb20c88c32d9b33fa57d5ed
+oid sha256:2455e70e00b8e172ba59d856d56c7f9c4d3d47824ed0b152214282ad8fea8176
 size 1155772233

checkpoint-2500/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9d683575abafeb1de186d6f13b563f85c30f963eb69b2647207f006dac8a4db2
+oid sha256:fe8b4067c8d3a7b38de39cea5b7ef489843427a694b5092a0845b6f6c0abe170
 size 14244

checkpoint-2500/trainer_state.json

@@ -1,7 +1,7 @@
 {
-  "best_metric": 0.4174647927284241,
-  "best_model_checkpoint": "mikhail_panzo/zlm_b64_le5_s12000/checkpoint-2500",
-  "epoch": 2.094240837696335,
+  "best_metric": 0.4208170473575592,
+  "best_model_checkpoint": "mikhail-panzo/zlm_b64_le5_s12000/checkpoint-2500",
+  "epoch": 2.0938023450586263,
   "eval_steps": 500,
   "global_step": 2500,
   "is_hyper_param_search": false,
@@ -9,393 +9,393 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.041884816753926704,
-      "grad_norm": 12.821951866149902,
-      "learning_rate": 2.4000000000000003e-07,
-      "loss": 1.1152,
+      "epoch": 0.04187604690117253,
+      "grad_norm": 4.788218975067139,
+      "learning_rate": 2.4500000000000004e-07,
+      "loss": 1.1819,
       "step": 50
     },
     {
-      "epoch": 0.08376963350785341,
-      "grad_norm": 6.186221122741699,
-      "learning_rate": 4.900000000000001e-07,
-      "loss": 1.0584,
+      "epoch": 0.08375209380234507,
+      "grad_norm": 15.618935585021973,
+      "learning_rate": 4.95e-07,
+      "loss": 1.0042,
       "step": 100
     },
     {
-      "epoch": 0.1256544502617801,
-      "grad_norm": 13.08674430847168,
+      "epoch": 0.12562814070351758,
+      "grad_norm": 2.7821555137634277,
       "learning_rate": 7.4e-07,
-      "loss": 0.9996,
+      "loss": 0.9676,
       "step": 150
     },
     {
-      "epoch": 0.16753926701570682,
-      "grad_norm": 4.10632848739624,
-      "learning_rate": 9.9e-07,
-      "loss": 0.9355,
+      "epoch": 0.16750418760469013,
+      "grad_norm": 3.759263038635254,
+      "learning_rate": 9.85e-07,
+      "loss": 0.9362,
       "step": 200
     },
     {
-      "epoch": 0.2094240837696335,
-      "grad_norm": 3.2428064346313477,
-      "learning_rate": 1.2400000000000002e-06,
-      "loss": 0.852,
+      "epoch": 0.20938023450586266,
+      "grad_norm": 3.906439781188965,
+      "learning_rate": 1.235e-06,
+      "loss": 0.8667,
       "step": 250
     },
     {
-      "epoch": 0.2513089005235602,
-      "grad_norm": 2.0861401557922363,
-      "learning_rate": 1.4900000000000001e-06,
-      "loss": 0.8241,
+      "epoch": 0.25125628140703515,
+      "grad_norm": 4.397643566131592,
+      "learning_rate": 1.485e-06,
+      "loss": 0.8455,
       "step": 300
     },
     {
-      "epoch": 0.2931937172774869,
-      "grad_norm": 8.602317810058594,
-      "learning_rate": 1.74e-06,
-      "loss": 0.7629,
+      "epoch": 0.2931323283082077,
+      "grad_norm": 2.94077467918396,
+      "learning_rate": 1.7350000000000001e-06,
+      "loss": 0.8376,
       "step": 350
     },
     {
-      "epoch": 0.33507853403141363,
-      "grad_norm": 2.7594990730285645,
-      "learning_rate": 1.9900000000000004e-06,
-      "loss": 0.7418,
+      "epoch": 0.33500837520938026,
+      "grad_norm": 4.161074161529541,
+      "learning_rate": 1.985e-06,
+      "loss": 0.7521,
       "step": 400
     },
     {
-      "epoch": 0.3769633507853403,
-      "grad_norm": 2.7047348022460938,
-      "learning_rate": 2.24e-06,
-      "loss": 0.7166,
+      "epoch": 0.3768844221105528,
+      "grad_norm": 2.8088552951812744,
+      "learning_rate": 2.235e-06,
+      "loss": 0.7444,
       "step": 450
     },
     {
-      "epoch": 0.418848167539267,
-      "grad_norm": 2.0666158199310303,
-      "learning_rate": 2.4900000000000003e-06,
-      "loss": 0.6876,
+      "epoch": 0.4187604690117253,
+      "grad_norm": 2.62283992767334,
+      "learning_rate": 2.4850000000000003e-06,
+      "loss": 0.7129,
       "step": 500
     },
     {
-      "epoch": 0.418848167539267,
-      "eval_loss": 0.5937426090240479,
-      "eval_runtime": 263.0955,
-      "eval_samples_per_second": 32.266,
-      "eval_steps_per_second": 4.037,
+      "epoch": 0.4187604690117253,
+      "eval_loss": 0.6066474914550781,
+      "eval_runtime": 213.4911,
+      "eval_samples_per_second": 39.763,
+      "eval_steps_per_second": 4.974,
       "step": 500
     },
     {
-      "epoch": 0.4607329842931937,
-      "grad_norm": 2.955193519592285,
-      "learning_rate": 2.7400000000000004e-06,
-      "loss": 0.673,
+      "epoch": 0.46063651591289784,
+      "grad_norm": 3.654665946960449,
+      "learning_rate": 2.7350000000000005e-06,
+      "loss": 0.6772,
       "step": 550
     },
     {
-      "epoch": 0.5026178010471204,
-      "grad_norm": 4.856322765350342,
-      "learning_rate": 2.99e-06,
-      "loss": 0.6539,
+      "epoch": 0.5025125628140703,
+      "grad_norm": 5.1851887702941895,
+      "learning_rate": 2.9850000000000002e-06,
+      "loss": 0.6552,
       "step": 600
     },
     {
-      "epoch": 0.5445026178010471,
-      "grad_norm": 7.631103992462158,
-      "learning_rate": 3.2400000000000003e-06,
-      "loss": 0.6588,
+      "epoch": 0.5443886097152428,
+      "grad_norm": 4.017187595367432,
+      "learning_rate": 3.2350000000000004e-06,
+      "loss": 0.6506,
       "step": 650
     },
     {
-      "epoch": 0.5863874345549738,
-      "grad_norm": 2.715771436691284,
-      "learning_rate": 3.49e-06,
-      "loss": 0.624,
+      "epoch": 0.5862646566164154,
+      "grad_norm": 2.1161563396453857,
+      "learning_rate": 3.485e-06,
+      "loss": 0.6301,
       "step": 700
     },
     {
-      "epoch": 0.6282722513089005,
-      "grad_norm": 2.7684273719787598,
-      "learning_rate": 3.74e-06,
-      "loss": 0.6116,
+      "epoch": 0.628140703517588,
+      "grad_norm": 1.857313871383667,
+      "learning_rate": 3.7350000000000002e-06,
+      "loss": 0.6373,
       "step": 750
     },
     {
-      "epoch": 0.6701570680628273,
-      "grad_norm": 2.219599723815918,
-      "learning_rate": 3.990000000000001e-06,
-      "loss": 0.6097,
+      "epoch": 0.6700167504187605,
+      "grad_norm": 3.6841330528259277,
+      "learning_rate": 3.985000000000001e-06,
+      "loss": 0.6305,
       "step": 800
     },
     {
-      "epoch": 0.7120418848167539,
-      "grad_norm": 3.749188184738159,
-      "learning_rate": 4.24e-06,
-      "loss": 0.614,
+      "epoch": 0.711892797319933,
+      "grad_norm": 2.6450912952423096,
+      "learning_rate": 4.235e-06,
+      "loss": 0.6169,
       "step": 850
     },
     {
-      "epoch": 0.7539267015706806,
-      "grad_norm": 2.120619773864746,
-      "learning_rate": 4.49e-06,
-      "loss": 0.5814,
+      "epoch": 0.7537688442211056,
+      "grad_norm": 1.7241184711456299,
+      "learning_rate": 4.485e-06,
+      "loss": 0.5933,
       "step": 900
     },
     {
-      "epoch": 0.7958115183246073,
-      "grad_norm": 2.7725327014923096,
-      "learning_rate": 4.74e-06,
-      "loss": 0.5895,
+      "epoch": 0.7956448911222781,
+      "grad_norm": 2.602735996246338,
+      "learning_rate": 4.735e-06,
+      "loss": 0.583,
       "step": 950
     },
     {
-      "epoch": 0.837696335078534,
-      "grad_norm": 9.851961135864258,
-      "learning_rate": 4.9900000000000005e-06,
-      "loss": 0.5623,
+      "epoch": 0.8375209380234506,
+      "grad_norm": 12.419028282165527,
+      "learning_rate": 4.9850000000000006e-06,
+      "loss": 0.5922,
       "step": 1000
     },
     {
-      "epoch": 0.837696335078534,
-      "eval_loss": 0.4933945834636688,
-      "eval_runtime": 258.739,
-      "eval_samples_per_second": 32.809,
-      "eval_steps_per_second": 4.105,
+      "epoch": 0.8375209380234506,
+      "eval_loss": 0.4979843497276306,
+      "eval_runtime": 214.0502,
+      "eval_samples_per_second": 39.659,
+      "eval_steps_per_second": 4.961,
       "step": 1000
     },
     {
-      "epoch": 0.8795811518324608,
-      "grad_norm": 5.074089527130127,
-      "learning_rate": 5.240000000000001e-06,
-      "loss": 0.5748,
+      "epoch": 0.8793969849246231,
+      "grad_norm": 3.200178861618042,
+      "learning_rate": 5.235e-06,
+      "loss": 0.5678,
       "step": 1050
     },
     {
-      "epoch": 0.9214659685863874,
-      "grad_norm": 5.884639739990234,
+      "epoch": 0.9212730318257957,
+      "grad_norm": 2.458869218826294,
       "learning_rate": 5.485e-06,
-      "loss": 0.5395,
+      "loss": 0.5558,
       "step": 1100
     },
     {
-      "epoch": 0.9633507853403142,
-      "grad_norm": 3.1588447093963623,
+      "epoch": 0.9631490787269682,
+      "grad_norm": 2.341752290725708,
       "learning_rate": 5.735e-06,
-      "loss": 0.5478,
+      "loss": 0.5666,
       "step": 1150
     },
     {
-      "epoch": 1.0052356020942408,
-      "grad_norm": 2.4452970027923584,
+      "epoch": 1.0050251256281406,
+      "grad_norm": 2.9191603660583496,
       "learning_rate": 5.985000000000001e-06,
-      "loss": 0.538,
+      "loss": 0.5345,
       "step": 1200
     },
     {
-      "epoch": 1.0471204188481675,
-      "grad_norm": 8.290769577026367,
+      "epoch": 1.0469011725293131,
+      "grad_norm": 2.68806529045105,
       "learning_rate": 6.235000000000001e-06,
-      "loss": 0.5411,
+      "loss": 0.5653,
       "step": 1250
     },
     {
-      "epoch": 1.0890052356020943,
-      "grad_norm": 4.080046653747559,
+      "epoch": 1.0887772194304857,
+      "grad_norm": 3.278820276260376,
       "learning_rate": 6.485000000000001e-06,
-      "loss": 0.5387,
+      "loss": 0.5565,
       "step": 1300
     },
     {
-      "epoch": 1.130890052356021,
-      "grad_norm": 2.7520201206207275,
+      "epoch": 1.1306532663316582,
+      "grad_norm": 3.066511869430542,
       "learning_rate": 6.735000000000001e-06,
-      "loss": 0.5426,
+      "loss": 0.5644,
       "step": 1350
     },
     {
-      "epoch": 1.1727748691099475,
-      "grad_norm": 3.8192107677459717,
+      "epoch": 1.1725293132328307,
+      "grad_norm": 4.269862651824951,
       "learning_rate": 6.985000000000001e-06,
-      "loss": 0.5315,
+      "loss": 0.5356,
       "step": 1400
     },
     {
-      "epoch": 1.2146596858638743,
-      "grad_norm": 9.535676956176758,
+      "epoch": 1.2144053601340032,
+      "grad_norm": 3.6681156158447266,
       "learning_rate": 7.235000000000001e-06,
-      "loss": 0.5159,
+      "loss": 0.5246,
       "step": 1450
     },
     {
-      "epoch": 1.256544502617801,
-      "grad_norm": 2.5731022357940674,
+      "epoch": 1.2562814070351758,
+      "grad_norm": 3.1227505207061768,
       "learning_rate": 7.485000000000001e-06,
-      "loss": 0.5165,
+      "loss": 0.5168,
       "step": 1500
     },
     {
-      "epoch": 1.256544502617801,
-      "eval_loss": 0.4628298878669739,
-      "eval_runtime": 257.0265,
-      "eval_samples_per_second": 33.028,
-      "eval_steps_per_second": 4.132,
+      "epoch": 1.2562814070351758,
+      "eval_loss": 0.45785218477249146,
+      "eval_runtime": 211.6749,
+      "eval_samples_per_second": 40.104,
+      "eval_steps_per_second": 5.017,
       "step": 1500
     },
     {
-      "epoch": 1.2984293193717278,
-      "grad_norm": 3.662090539932251,
+      "epoch": 1.2981574539363483,
+      "grad_norm": 2.6239662170410156,
       "learning_rate": 7.735e-06,
-      "loss": 0.5188,
+      "loss": 0.5153,
       "step": 1550
     },
     {
-      "epoch": 1.3403141361256545,
-      "grad_norm": 3.1680572032928467,
+      "epoch": 1.3400335008375208,
+      "grad_norm": 2.881901979446411,
       "learning_rate": 7.985e-06,
-      "loss": 0.5148,
+      "loss": 0.518,
       "step": 1600
     },
     {
-      "epoch": 1.3821989528795813,
-      "grad_norm": 3.601471185684204,
+      "epoch": 1.3819095477386933,
+      "grad_norm": 2.8052310943603516,
       "learning_rate": 8.235e-06,
-      "loss": 0.5127,
+      "loss": 0.5272,
       "step": 1650
     },
     {
-      "epoch": 1.4240837696335078,
-      "grad_norm": 4.497046947479248,
+      "epoch": 1.4237855946398659,
+      "grad_norm": 2.6334052085876465,
       "learning_rate": 8.485000000000001e-06,
-      "loss": 0.5039,
+      "loss": 0.5188,
       "step": 1700
     },
     {
-      "epoch": 1.4659685863874345,
-      "grad_norm": 3.6496424674987793,
+      "epoch": 1.4656616415410384,
+      "grad_norm": 3.038274049758911,
       "learning_rate": 8.735000000000002e-06,
-      "loss": 0.5114,
+      "loss": 0.5139,
       "step": 1750
     },
     {
-      "epoch": 1.5078534031413613,
-      "grad_norm": 2.7080061435699463,
+      "epoch": 1.507537688442211,
+      "grad_norm": 3.2438852787017822,
       "learning_rate": 8.985000000000001e-06,
-      "loss": 0.4953,
+      "loss": 0.5092,
       "step": 1800
     },
     {
-      "epoch": 1.5497382198952878,
-      "grad_norm": 2.5439274311065674,
+      "epoch": 1.5494137353433834,
+      "grad_norm": 4.168883800506592,
       "learning_rate": 9.235e-06,
-      "loss": 0.497,
+      "loss": 0.5103,
       "step": 1850
     },
     {
-      "epoch": 1.5916230366492146,
-      "grad_norm": 2.004573106765747,
+      "epoch": 1.591289782244556,
+      "grad_norm": 2.6399967670440674,
       "learning_rate": 9.485000000000002e-06,
-      "loss": 0.4889,
+      "loss": 0.5025,
       "step": 1900
     },
     {
-      "epoch": 1.6335078534031413,
-      "grad_norm": 4.699184894561768,
+      "epoch": 1.6331658291457285,
+      "grad_norm": 1.6828950643539429,
       "learning_rate": 9.735e-06,
-      "loss": 0.4925,
+      "loss": 0.4963,
       "step": 1950
     },
     {
-      "epoch": 1.675392670157068,
-      "grad_norm": 3.2080323696136475,
+      "epoch": 1.675041876046901,
+      "grad_norm": 1.8955535888671875,
       "learning_rate": 9.985000000000002e-06,
-      "loss": 0.4851,
+      "loss": 0.4954,
       "step": 2000
     },
     {
-      "epoch": 1.675392670157068,
-      "eval_loss": 0.43240800499916077,
-      "eval_runtime": 256.4745,
-      "eval_samples_per_second": 33.099,
-      "eval_steps_per_second": 4.141,
+      "epoch": 1.675041876046901,
+      "eval_loss": 0.43827974796295166,
+      "eval_runtime": 212.1455,
+      "eval_samples_per_second": 40.015,
+      "eval_steps_per_second": 5.006,
       "step": 2000
     },
     {
-      "epoch": 1.7172774869109948,
-      "grad_norm": 2.5463037490844727,
+      "epoch": 1.7169179229480735,
+      "grad_norm": 3.9634511470794678,
       "learning_rate": 9.953000000000001e-06,
-      "loss": 0.4802,
+      "loss": 0.4891,
       "step": 2050
     },
     {
-      "epoch": 1.7591623036649215,
-      "grad_norm": 3.6697189807891846,
+      "epoch": 1.758793969849246,
+      "grad_norm": 3.5446126461029053,
       "learning_rate": 9.903e-06,
-      "loss": 0.485,
+      "loss": 0.5141,
       "step": 2100
     },
     {
-      "epoch": 1.8010471204188483,
-      "grad_norm": 2.6455466747283936,
+      "epoch": 1.8006700167504186,
+      "grad_norm": 3.1621015071868896,
       "learning_rate": 9.853e-06,
-      "loss": 0.4863,
+      "loss": 0.4941,
       "step": 2150
     },
     {
-      "epoch": 1.8429319371727748,
-      "grad_norm": 2.849780321121216,
+      "epoch": 1.8425460636515911,
+      "grad_norm": 2.6622440814971924,
       "learning_rate": 9.803e-06,
-      "loss": 0.4824,
+      "loss": 0.4979,
       "step": 2200
     },
     {
-      "epoch": 1.8848167539267016,
-      "grad_norm": 2.037757158279419,
+      "epoch": 1.8844221105527639,
+      "grad_norm": 1.7774969339370728,
       "learning_rate": 9.753e-06,
-      "loss": 0.4762,
+      "loss": 0.4768,
       "step": 2250
     },
     {
-      "epoch": 1.9267015706806283,
-      "grad_norm": 2.060755491256714,
+      "epoch": 1.9262981574539364,
+      "grad_norm": 2.1141610145568848,
       "learning_rate": 9.703000000000002e-06,
-      "loss": 0.4793,
+      "loss": 0.4834,
       "step": 2300
     },
     {
-      "epoch": 1.9685863874345548,
-      "grad_norm": 2.3721907138824463,
+      "epoch": 1.968174204355109,
+      "grad_norm": 3.4576170444488525,
       "learning_rate": 9.653e-06,
-      "loss": 0.4744,
+      "loss": 0.493,
       "step": 2350
     },
     {
-      "epoch": 2.0104712041884816,
-      "grad_norm": 2.605724573135376,
+      "epoch": 2.0100502512562812,
+      "grad_norm": 2.4238383769989014,
       "learning_rate": 9.603000000000001e-06,
-      "loss": 0.4665,
+      "loss": 0.4836,
       "step": 2400
     },
     {
-      "epoch": 2.0523560209424083,
-      "grad_norm": 1.9419960975646973,
+      "epoch": 2.0519262981574538,
+      "grad_norm": 2.2526774406433105,
       "learning_rate": 9.553000000000002e-06,
-      "loss": 0.4701,
+      "loss": 0.4832,
       "step": 2450
     },
     {
-      "epoch": 2.094240837696335,
-      "grad_norm": 2.3074657917022705,
+      "epoch": 2.0938023450586263,
+      "grad_norm": 2.8719518184661865,
       "learning_rate": 9.503e-06,
-      "loss": 0.4672,
+      "loss": 0.4836,
       "step": 2500
     },
     {
-      "epoch": 2.094240837696335,
-      "eval_loss": 0.4174647927284241,
-      "eval_runtime": 259.1813,
-      "eval_samples_per_second": 32.753,
-      "eval_steps_per_second": 4.098,
+      "epoch": 2.0938023450586263,
+      "eval_loss": 0.4208170473575592,
+      "eval_runtime": 214.437,
+      "eval_samples_per_second": 39.587,
+      "eval_steps_per_second": 4.953,
       "step": 2500
     }
   ],
@@ -416,8 +416,8 @@
       "attributes": {}
     }
   },
-  "total_flos": 2.2497608619313344e+16,
-  "train_batch_size": 16,
+  "total_flos": 2.436267617348717e+16,
+  "train_batch_size": 32,
   "trial_name": null,
   "trial_params": null
 }

checkpoint-2500/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b20f3b06aa67136878a19fd6f17a63e91db785ca54662fa7debf3d46f1fe910d
+oid sha256:f4687cdb2b9124db8f1d7d7525bb9403c3df24fb82ca7d74640c10781e1e94be
 size 5304