Training in progress, step 2500, checkpoint

checkpoint-2500/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e953d290bf6fa77196bebdccebcf1449321dfb1e0785c52fad08116743eaa121
+oid sha256:e30140663f854ebc5662f20e25fe19ee61a374e2a8d3b215ab985aafc932d680
 size 577789320

checkpoint-2500/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ac921c6e5b5baf94cb84864ccefe951eb782be9d911539577b7f87581a2f2238
+oid sha256:8c413f0043ffce4cf67002924d4e0d28784d5357af9c310ca3c2190feb8b57e4
 size 1155772233

checkpoint-2500/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17d91fb6e687c9e485a8b0f5b25e3d3502d313b85fb6cd013a370aa3ac81e5d9
+oid sha256:f593e5b8cc8b099c125a8ea02d311833e867baf718b249e576a84f665a3b9893
 size 14244

checkpoint-2500/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:01c8f493f88d64182a2dfcee190fa9ab01bd14fa7f93a04b41d568c47a164cf9
+oid sha256:0066d344f9ec09056d79be503a6fa7d2168a78b6133e8a24c854cb818188b1c7
 size 1064

checkpoint-2500/trainer_state.json

@@ -1,7 +1,7 @@
 {
-  "best_metric": 0.35975897312164307,
+  "best_metric": 0.3701198697090149,
   "best_model_checkpoint": "mikhail_panzo/zlm_b64_le4_s12000/checkpoint-2500",
-  "epoch": 2.094869109947644,
+  "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": 2.697540283203125,
-      "learning_rate": 2.4000000000000003e-06,
-      "loss": 1.0526,
+      "epoch": 0.04187604690117253,
+      "grad_norm": 3.0184223651885986,
+      "learning_rate": 2.35e-06,
+      "loss": 1.1206,
       "step": 50
     },
     {
-      "epoch": 0.08376963350785341,
-      "grad_norm": 4.0351996421813965,
-      "learning_rate": 4.9000000000000005e-06,
-      "loss": 0.8484,
+      "epoch": 0.08375209380234507,
+      "grad_norm": 3.4442005157470703,
+      "learning_rate": 4.85e-06,
+      "loss": 0.8185,
       "step": 100
     },
     {
-      "epoch": 0.1256544502617801,
-      "grad_norm": 4.858840465545654,
-      "learning_rate": 7.4e-06,
-      "loss": 0.7505,
+      "epoch": 0.12562814070351758,
+      "grad_norm": 2.6812493801116943,
+      "learning_rate": 7.35e-06,
+      "loss": 0.7576,
       "step": 150
     },
     {
-      "epoch": 0.16753926701570682,
-      "grad_norm": 2.396157741546631,
-      "learning_rate": 9.900000000000002e-06,
-      "loss": 0.6889,
+      "epoch": 0.16750418760469013,
+      "grad_norm": 4.518482685089111,
+      "learning_rate": 9.85e-06,
+      "loss": 0.6938,
       "step": 200
     },
     {
-      "epoch": 0.2094240837696335,
-      "grad_norm": 2.660891056060791,
-      "learning_rate": 1.24e-05,
-      "loss": 0.6363,
+      "epoch": 0.20938023450586266,
+      "grad_norm": 4.683146953582764,
+      "learning_rate": 1.235e-05,
+      "loss": 0.6386,
       "step": 250
     },
     {
-      "epoch": 0.2513089005235602,
-      "grad_norm": 2.848365068435669,
-      "learning_rate": 1.49e-05,
-      "loss": 0.6107,
+      "epoch": 0.25125628140703515,
+      "grad_norm": 2.7392578125,
+      "learning_rate": 1.485e-05,
+      "loss": 0.6226,
       "step": 300
     },
     {
-      "epoch": 0.2931937172774869,
-      "grad_norm": 2.0031681060791016,
-      "learning_rate": 1.74e-05,
-      "loss": 0.5673,
+      "epoch": 0.2931323283082077,
+      "grad_norm": 2.017056465148926,
+      "learning_rate": 1.7349999999999998e-05,
+      "loss": 0.6128,
       "step": 350
     },
     {
-      "epoch": 0.33507853403141363,
-      "grad_norm": 4.771584510803223,
-      "learning_rate": 1.9900000000000003e-05,
-      "loss": 0.5566,
+      "epoch": 0.33500837520938026,
+      "grad_norm": 3.6669230461120605,
+      "learning_rate": 1.985e-05,
+      "loss": 0.5675,
       "step": 400
     },
     {
-      "epoch": 0.3769633507853403,
-      "grad_norm": 9.876974105834961,
-      "learning_rate": 2.2400000000000002e-05,
-      "loss": 0.5487,
+      "epoch": 0.3768844221105528,
+      "grad_norm": 4.16896915435791,
+      "learning_rate": 2.235e-05,
+      "loss": 0.5655,
       "step": 450
     },
     {
-      "epoch": 0.418848167539267,
-      "grad_norm": 2.113546848297119,
-      "learning_rate": 2.4900000000000002e-05,
-      "loss": 0.5268,
+      "epoch": 0.4187604690117253,
+      "grad_norm": 3.691459894180298,
+      "learning_rate": 2.485e-05,
+      "loss": 0.5487,
       "step": 500
     },
     {
-      "epoch": 0.418848167539267,
-      "eval_loss": 0.46557939052581787,
-      "eval_runtime": 297.9193,
-      "eval_samples_per_second": 28.494,
-      "eval_steps_per_second": 3.565,
+      "epoch": 0.4187604690117253,
+      "eval_loss": 0.47458764910697937,
+      "eval_runtime": 214.7196,
+      "eval_samples_per_second": 39.535,
+      "eval_steps_per_second": 4.946,
       "step": 500
     },
     {
-      "epoch": 0.4607329842931937,
-      "grad_norm": 2.7003655433654785,
-      "learning_rate": 2.7400000000000002e-05,
-      "loss": 0.5212,
+      "epoch": 0.46063651591289784,
+      "grad_norm": 3.3792996406555176,
+      "learning_rate": 2.7350000000000004e-05,
+      "loss": 0.5232,
       "step": 550
     },
     {
-      "epoch": 0.5026178010471204,
-      "grad_norm": 5.649191379547119,
-      "learning_rate": 2.9900000000000002e-05,
-      "loss": 0.5138,
+      "epoch": 0.5025125628140703,
+      "grad_norm": 3.6263363361358643,
+      "learning_rate": 2.985e-05,
+      "loss": 0.5132,
       "step": 600
     },
     {
-      "epoch": 0.5445026178010471,
-      "grad_norm": 2.272181749343872,
-      "learning_rate": 3.24e-05,
-      "loss": 0.5121,
+      "epoch": 0.5443886097152428,
+      "grad_norm": 2.652538776397705,
+      "learning_rate": 3.235e-05,
+      "loss": 0.5067,
       "step": 650
     },
     {
-      "epoch": 0.5863874345549738,
-      "grad_norm": 2.957671642303467,
-      "learning_rate": 3.49e-05,
-      "loss": 0.496,
+      "epoch": 0.5862646566164154,
+      "grad_norm": 3.5355172157287598,
+      "learning_rate": 3.485e-05,
+      "loss": 0.4981,
       "step": 700
     },
     {
-      "epoch": 0.6282722513089005,
-      "grad_norm": 3.577693462371826,
-      "learning_rate": 3.74e-05,
-      "loss": 0.4886,
+      "epoch": 0.628140703517588,
+      "grad_norm": 2.5035619735717773,
+      "learning_rate": 3.735e-05,
+      "loss": 0.5094,
       "step": 750
     },
     {
-      "epoch": 0.6701570680628273,
-      "grad_norm": 2.466113328933716,
-      "learning_rate": 3.99e-05,
-      "loss": 0.4879,
+      "epoch": 0.6700167504187605,
+      "grad_norm": 2.9585518836975098,
+      "learning_rate": 3.9850000000000006e-05,
+      "loss": 0.5053,
       "step": 800
     },
     {
-      "epoch": 0.7120418848167539,
-      "grad_norm": 2.941208600997925,
-      "learning_rate": 4.24e-05,
-      "loss": 0.4865,
+      "epoch": 0.711892797319933,
+      "grad_norm": 9.997116088867188,
+      "learning_rate": 4.235e-05,
+      "loss": 0.502,
       "step": 850
     },
     {
-      "epoch": 0.7539267015706806,
-      "grad_norm": 2.3843142986297607,
-      "learning_rate": 4.49e-05,
-      "loss": 0.4689,
+      "epoch": 0.7537688442211056,
+      "grad_norm": 1.8630377054214478,
+      "learning_rate": 4.4850000000000006e-05,
+      "loss": 0.483,
       "step": 900
     },
     {
-      "epoch": 0.7958115183246073,
-      "grad_norm": 2.232208728790283,
-      "learning_rate": 4.74e-05,
-      "loss": 0.4768,
+      "epoch": 0.7956448911222781,
+      "grad_norm": 1.5534099340438843,
+      "learning_rate": 4.735e-05,
+      "loss": 0.4743,
       "step": 950
     },
     {
-      "epoch": 0.837696335078534,
-      "grad_norm": 2.1702969074249268,
-      "learning_rate": 4.99e-05,
-      "loss": 0.4583,
+      "epoch": 0.8375209380234506,
+      "grad_norm": 2.8155503273010254,
+      "learning_rate": 4.9850000000000006e-05,
+      "loss": 0.483,
       "step": 1000
     },
     {
-      "epoch": 0.837696335078534,
-      "eval_loss": 0.4044812321662903,
-      "eval_runtime": 296.3081,
-      "eval_samples_per_second": 28.649,
-      "eval_steps_per_second": 3.584,
+      "epoch": 0.8375209380234506,
+      "eval_loss": 0.4226582944393158,
+      "eval_runtime": 213.2339,
+      "eval_samples_per_second": 39.811,
+      "eval_steps_per_second": 4.98,
       "step": 1000
     },
     {
-      "epoch": 0.8795811518324608,
-      "grad_norm": 2.7613942623138428,
-      "learning_rate": 5.2400000000000007e-05,
-      "loss": 0.4664,
+      "epoch": 0.8793969849246231,
+      "grad_norm": 4.2074055671691895,
+      "learning_rate": 5.235e-05,
+      "loss": 0.4625,
       "step": 1050
     },
     {
-      "epoch": 0.9214659685863874,
-      "grad_norm": 2.4752321243286133,
-      "learning_rate": 5.4900000000000006e-05,
-      "loss": 0.442,
+      "epoch": 0.9212730318257957,
+      "grad_norm": 2.215416431427002,
+      "learning_rate": 5.485e-05,
+      "loss": 0.461,
       "step": 1100
     },
     {
-      "epoch": 0.9633507853403142,
-      "grad_norm": 2.27407169342041,
-      "learning_rate": 5.74e-05,
-      "loss": 0.4498,
+      "epoch": 0.9631490787269682,
+      "grad_norm": 3.3321831226348877,
+      "learning_rate": 5.7350000000000005e-05,
+      "loss": 0.4807,
       "step": 1150
     },
     {
-      "epoch": 1.0052356020942408,
-      "grad_norm": 1.8998862504959106,
-      "learning_rate": 5.99e-05,
-      "loss": 0.4433,
+      "epoch": 1.0050251256281406,
+      "grad_norm": 3.5796420574188232,
+      "learning_rate": 5.9850000000000005e-05,
+      "loss": 0.4456,
       "step": 1200
     },
     {
-      "epoch": 1.0471204188481675,
-      "grad_norm": 2.8479738235473633,
-      "learning_rate": 6.24e-05,
-      "loss": 0.4417,
+      "epoch": 1.0469011725293131,
+      "grad_norm": 2.3565518856048584,
+      "learning_rate": 6.235000000000001e-05,
+      "loss": 0.4664,
       "step": 1250
     },
     {
-      "epoch": 1.0890052356020943,
-      "grad_norm": 1.9095953702926636,
-      "learning_rate": 6.49e-05,
-      "loss": 0.444,
+      "epoch": 1.0887772194304857,
+      "grad_norm": 3.147794485092163,
+      "learning_rate": 6.485e-05,
+      "loss": 0.4681,
       "step": 1300
     },
     {
-      "epoch": 1.130890052356021,
-      "grad_norm": 1.368077278137207,
-      "learning_rate": 6.740000000000001e-05,
-      "loss": 0.4519,
+      "epoch": 1.1306532663316582,
+      "grad_norm": 3.028409957885742,
+      "learning_rate": 6.735e-05,
+      "loss": 0.4663,
       "step": 1350
     },
     {
-      "epoch": 1.1727748691099475,
-      "grad_norm": 2.0456721782684326,
-      "learning_rate": 6.99e-05,
-      "loss": 0.4411,
+      "epoch": 1.1725293132328307,
+      "grad_norm": 4.820915222167969,
+      "learning_rate": 6.985e-05,
+      "loss": 0.4493,
       "step": 1400
     },
     {
-      "epoch": 1.2146596858638743,
-      "grad_norm": 1.6217619180679321,
-      "learning_rate": 7.24e-05,
-      "loss": 0.4324,
+      "epoch": 1.2144053601340032,
+      "grad_norm": 1.93254816532135,
+      "learning_rate": 7.235000000000001e-05,
+      "loss": 0.4413,
       "step": 1450
     },
     {
-      "epoch": 1.256544502617801,
-      "grad_norm": 2.961646795272827,
-      "learning_rate": 7.49e-05,
-      "loss": 0.4321,
+      "epoch": 1.2562814070351758,
+      "grad_norm": 2.237015724182129,
+      "learning_rate": 7.485e-05,
+      "loss": 0.432,
       "step": 1500
     },
     {
-      "epoch": 1.256544502617801,
-      "eval_loss": 0.3930450975894928,
-      "eval_runtime": 301.1428,
-      "eval_samples_per_second": 28.189,
-      "eval_steps_per_second": 3.527,
+      "epoch": 1.2562814070351758,
+      "eval_loss": 0.39829039573669434,
+      "eval_runtime": 217.3782,
+      "eval_samples_per_second": 39.052,
+      "eval_steps_per_second": 4.885,
       "step": 1500
     },
     {
-      "epoch": 1.2984293193717278,
-      "grad_norm": 2.543201446533203,
-      "learning_rate": 7.740000000000001e-05,
-      "loss": 0.4364,
+      "epoch": 1.2981574539363483,
+      "grad_norm": 1.709120273590088,
+      "learning_rate": 7.735e-05,
+      "loss": 0.4361,
       "step": 1550
     },
     {
-      "epoch": 1.3403141361256545,
-      "grad_norm": 1.5807560682296753,
-      "learning_rate": 7.99e-05,
-      "loss": 0.433,
+      "epoch": 1.3400335008375208,
+      "grad_norm": 3.733328104019165,
+      "learning_rate": 7.985e-05,
+      "loss": 0.4357,
       "step": 1600
     },
     {
-      "epoch": 1.3821989528795813,
-      "grad_norm": 1.587385892868042,
-      "learning_rate": 8.24e-05,
-      "loss": 0.4285,
+      "epoch": 1.3819095477386933,
+      "grad_norm": 4.131324768066406,
+      "learning_rate": 8.235000000000001e-05,
+      "loss": 0.4445,
       "step": 1650
     },
     {
-      "epoch": 1.4240837696335078,
-      "grad_norm": 2.1177687644958496,
-      "learning_rate": 8.49e-05,
-      "loss": 0.4209,
+      "epoch": 1.4237855946398659,
+      "grad_norm": 2.2893927097320557,
+      "learning_rate": 8.485e-05,
+      "loss": 0.4379,
       "step": 1700
     },
     {
-      "epoch": 1.4659685863874345,
-      "grad_norm": 1.6078702211380005,
-      "learning_rate": 8.740000000000001e-05,
-      "loss": 0.426,
+      "epoch": 1.4656616415410384,
+      "grad_norm": 2.1464734077453613,
+      "learning_rate": 8.735000000000001e-05,
+      "loss": 0.4393,
       "step": 1750
     },
     {
-      "epoch": 1.5078534031413613,
-      "grad_norm": 2.10492205619812,
-      "learning_rate": 8.99e-05,
-      "loss": 0.4191,
+      "epoch": 1.507537688442211,
+      "grad_norm": 2.01698637008667,
+      "learning_rate": 8.985e-05,
+      "loss": 0.4341,
       "step": 1800
     },
     {
-      "epoch": 1.5497382198952878,
-      "grad_norm": 2.169309616088867,
-      "learning_rate": 9.240000000000001e-05,
-      "loss": 0.4178,
+      "epoch": 1.5494137353433834,
+      "grad_norm": 2.259115219116211,
+      "learning_rate": 9.235000000000001e-05,
+      "loss": 0.4324,
       "step": 1850
     },
     {
-      "epoch": 1.5916230366492146,
-      "grad_norm": 1.559006929397583,
-      "learning_rate": 9.49e-05,
-      "loss": 0.4133,
+      "epoch": 1.591289782244556,
+      "grad_norm": 2.0072858333587646,
+      "learning_rate": 9.485e-05,
+      "loss": 0.4327,
       "step": 1900
     },
     {
-      "epoch": 1.6335078534031413,
-      "grad_norm": 2.0436346530914307,
-      "learning_rate": 9.74e-05,
-      "loss": 0.4234,
+      "epoch": 1.6331658291457285,
+      "grad_norm": 2.060303211212158,
+      "learning_rate": 9.735000000000001e-05,
+      "loss": 0.4279,
       "step": 1950
     },
     {
-      "epoch": 1.675392670157068,
-      "grad_norm": 2.089486837387085,
-      "learning_rate": 9.99e-05,
-      "loss": 0.4092,
+      "epoch": 1.675041876046901,
+      "grad_norm": 1.2846741676330566,
+      "learning_rate": 9.985000000000001e-05,
+      "loss": 0.429,
       "step": 2000
     },
     {
-      "epoch": 1.675392670157068,
-      "eval_loss": 0.37686464190483093,
-      "eval_runtime": 310.3113,
-      "eval_samples_per_second": 27.356,
-      "eval_steps_per_second": 3.422,
+      "epoch": 1.675041876046901,
+      "eval_loss": 0.3953319191932678,
+      "eval_runtime": 217.8108,
+      "eval_samples_per_second": 38.974,
+      "eval_steps_per_second": 4.876,
       "step": 2000
     },
     {
-      "epoch": 1.7179057591623037,
-      "grad_norm": 2.684563398361206,
-      "learning_rate": 9.952e-05,
-      "loss": 0.4099,
+      "epoch": 1.7169179229480735,
+      "grad_norm": 2.253840923309326,
+      "learning_rate": 9.953e-05,
+      "loss": 0.4264,
       "step": 2050
     },
     {
-      "epoch": 1.7597905759162304,
-      "grad_norm": 1.7815254926681519,
-      "learning_rate": 9.902e-05,
-      "loss": 0.4093,
+      "epoch": 1.758793969849246,
+      "grad_norm": 1.3114299774169922,
+      "learning_rate": 9.903e-05,
+      "loss": 0.4355,
       "step": 2100
     },
     {
-      "epoch": 1.8016753926701572,
-      "grad_norm": 1.9900892972946167,
-      "learning_rate": 9.852e-05,
-      "loss": 0.4116,
+      "epoch": 1.8006700167504186,
+      "grad_norm": 1.8178610801696777,
+      "learning_rate": 9.853e-05,
+      "loss": 0.4256,
       "step": 2150
     },
     {
-      "epoch": 1.8435602094240837,
-      "grad_norm": 1.8637460470199585,
-      "learning_rate": 9.802e-05,
-      "loss": 0.4089,
+      "epoch": 1.8425460636515911,
+      "grad_norm": 2.1193735599517822,
+      "learning_rate": 9.803e-05,
+      "loss": 0.4305,
       "step": 2200
     },
     {
-      "epoch": 1.8854450261780105,
-      "grad_norm": 1.58171546459198,
-      "learning_rate": 9.752e-05,
-      "loss": 0.4053,
+      "epoch": 1.8844221105527639,
+      "grad_norm": 1.6871263980865479,
+      "learning_rate": 9.753e-05,
+      "loss": 0.4106,
       "step": 2250
     },
     {
-      "epoch": 1.9273298429319372,
-      "grad_norm": 1.4995630979537964,
-      "learning_rate": 9.702e-05,
-      "loss": 0.4049,
+      "epoch": 1.9262981574539364,
+      "grad_norm": 1.2699921131134033,
+      "learning_rate": 9.703000000000001e-05,
+      "loss": 0.4156,
       "step": 2300
     },
     {
-      "epoch": 1.9692146596858637,
-      "grad_norm": 2.447049379348755,
-      "learning_rate": 9.652e-05,
-      "loss": 0.4029,
+      "epoch": 1.968174204355109,
+      "grad_norm": 1.7966769933700562,
+      "learning_rate": 9.653000000000001e-05,
+      "loss": 0.4257,
       "step": 2350
     },
     {
-      "epoch": 2.0110994764397905,
-      "grad_norm": 1.8776074647903442,
-      "learning_rate": 9.602e-05,
-      "loss": 0.3988,
+      "epoch": 2.0100502512562812,
+      "grad_norm": 1.7838683128356934,
+      "learning_rate": 9.603000000000001e-05,
+      "loss": 0.4156,
       "step": 2400
     },
     {
-      "epoch": 2.0529842931937172,
-      "grad_norm": 3.1785056591033936,
-      "learning_rate": 9.552000000000001e-05,
-      "loss": 0.4023,
+      "epoch": 2.0519262981574538,
+      "grad_norm": 1.4161527156829834,
+      "learning_rate": 9.553e-05,
+      "loss": 0.4096,
       "step": 2450
     },
     {
-      "epoch": 2.094869109947644,
-      "grad_norm": 1.4416707754135132,
-      "learning_rate": 9.502000000000001e-05,
-      "loss": 0.4008,
+      "epoch": 2.0938023450586263,
+      "grad_norm": 1.5864896774291992,
+      "learning_rate": 9.503e-05,
+      "loss": 0.4168,
       "step": 2500
     },
     {
-      "epoch": 2.094869109947644,
-      "eval_loss": 0.35975897312164307,
-      "eval_runtime": 275.2855,
-      "eval_samples_per_second": 30.837,
-      "eval_steps_per_second": 3.858,
+      "epoch": 2.0938023450586263,
+      "eval_loss": 0.3701198697090149,
+      "eval_runtime": 220.4399,
+      "eval_samples_per_second": 38.509,
+      "eval_steps_per_second": 4.818,
       "step": 2500
     }
   ],
@@ -416,8 +416,8 @@
       "attributes": {}
     }
   },
-  "total_flos": 2.245101969215635e+16,
-  "train_batch_size": 16,
+  "total_flos": 2.431752842670624e+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:fa8f41edfc5db87ff9f21c8561c4e1b9e9e17c2b5ee34213317807a5e0e41371
+oid sha256:fb8203a5324b57ae1a64f032a5185016d65cbc7738fd0978239bc5b2518d051b
 size 5304