Training in progress, step 2500, checkpoint

checkpoint-2500/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:34108353347578833ac324969b4e9dc566c790cdf8b003c75b5e1fb6d0a08a2f
+oid sha256:b3f84d011e01713fdd7cc2462fc02b88b990d4282673e0f1f305c7c62a0e4b6e
 size 577789320

checkpoint-2500/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:641b731402942692aa155564b5f696d1a7f02c0d65f5aedc78c3ced20fb0843d
+oid sha256:94e89bcd33a6d50c7607315375ded70762b3e71fb3f9ce84f79fba6cea3c3b3f
 size 1155772233

checkpoint-2500/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a93dd6257561e3ec08839fe73c1f9f14f8e256595ebab46f2789d7bc17638c5f
+oid sha256:aa82a28e87976cd817afec1fe2e29654138caffccb5ad09dcd107e73988fc4fc
 size 14244

checkpoint-2500/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ba68065c0ba5863ea7a5974f2a119aa320d107d98a915a0d39227618f3cb2905
+oid sha256:a3aa360e51b0a30eb67391490a7cfbf76d08e435f0d7cd7c6b7dd9bf99bca988
 size 1064

checkpoint-2500/trainer_state.json

@@ -1,7 +1,7 @@
 {
-  "best_metric": 0.41223829984664917,
-  "best_model_checkpoint": "mikhail-panzo/zlm-fil-ceb_b64_le5_s8000/checkpoint-1500",
-  "epoch": 108.69565217391305,
+  "best_metric": 0.39545077085494995,
+  "best_model_checkpoint": "mikhail-panzo/zlm-fil-ceb_b64_le5_s8000/checkpoint-2500",
+  "epoch": 99.00990099009901,
   "eval_steps": 500,
   "global_step": 2500,
   "is_hyper_param_search": false,
@@ -9,400 +9,400 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 2.1739130434782608,
-      "grad_norm": 0.7357823252677917,
+      "epoch": 1.9801980198019802,
+      "grad_norm": 1.2683665752410889,
       "learning_rate": 2.5000000000000004e-07,
-      "loss": 0.4055,
+      "loss": 0.4932,
       "step": 50
     },
     {
-      "epoch": 4.3478260869565215,
-      "grad_norm": 0.7944665551185608,
+      "epoch": 3.9603960396039604,
+      "grad_norm": 1.1266463994979858,
       "learning_rate": 5.000000000000001e-07,
-      "loss": 0.3988,
+      "loss": 0.4898,
       "step": 100
     },
     {
-      "epoch": 6.521739130434782,
-      "grad_norm": 0.7226569652557373,
+      "epoch": 5.9405940594059405,
+      "grad_norm": 1.1800968647003174,
       "learning_rate": 7.5e-07,
-      "loss": 0.4072,
+      "loss": 0.4839,
       "step": 150
     },
     {
-      "epoch": 8.695652173913043,
-      "grad_norm": 0.7845238447189331,
+      "epoch": 7.920792079207921,
+      "grad_norm": 0.8235568404197693,
       "learning_rate": 1.0000000000000002e-06,
-      "loss": 0.4028,
+      "loss": 0.4785,
       "step": 200
     },
     {
-      "epoch": 10.869565217391305,
-      "grad_norm": 0.9769566655158997,
+      "epoch": 9.900990099009901,
+      "grad_norm": 1.313211441040039,
       "learning_rate": 1.25e-06,
-      "loss": 0.4066,
+      "loss": 0.4767,
       "step": 250
     },
     {
-      "epoch": 13.043478260869565,
-      "grad_norm": 0.8634074330329895,
+      "epoch": 11.881188118811881,
+      "grad_norm": 0.7831560373306274,
       "learning_rate": 1.5e-06,
-      "loss": 0.405,
+      "loss": 0.4681,
       "step": 300
     },
     {
-      "epoch": 15.217391304347826,
-      "grad_norm": 0.9029327034950256,
+      "epoch": 13.861386138613861,
+      "grad_norm": 0.7987237572669983,
       "learning_rate": 1.75e-06,
-      "loss": 0.4074,
+      "loss": 0.4658,
       "step": 350
     },
     {
-      "epoch": 17.391304347826086,
-      "grad_norm": 0.7624074220657349,
+      "epoch": 15.841584158415841,
+      "grad_norm": 0.7143813371658325,
       "learning_rate": 2.0000000000000003e-06,
-      "loss": 0.406,
+      "loss": 0.4627,
       "step": 400
     },
     {
-      "epoch": 19.565217391304348,
-      "grad_norm": 1.005196213722229,
+      "epoch": 17.821782178217823,
+      "grad_norm": 0.8037531971931458,
       "learning_rate": 2.25e-06,
-      "loss": 0.4052,
+      "loss": 0.461,
       "step": 450
     },
     {
-      "epoch": 21.73913043478261,
-      "grad_norm": 1.0276380777359009,
+      "epoch": 19.801980198019802,
+      "grad_norm": 0.79031902551651,
       "learning_rate": 2.5e-06,
-      "loss": 0.4019,
+      "loss": 0.4592,
       "step": 500
     },
     {
-      "epoch": 21.73913043478261,
-      "eval_loss": 0.4143177568912506,
-      "eval_runtime": 6.4074,
-      "eval_samples_per_second": 24.815,
-      "eval_steps_per_second": 3.121,
+      "epoch": 19.801980198019802,
+      "eval_loss": 0.42528408765792847,
+      "eval_runtime": 7.8982,
+      "eval_samples_per_second": 22.79,
+      "eval_steps_per_second": 2.912,
       "step": 500
     },
     {
-      "epoch": 23.91304347826087,
-      "grad_norm": 0.9067140817642212,
+      "epoch": 21.782178217821784,
+      "grad_norm": 0.7966288328170776,
       "learning_rate": 2.7500000000000004e-06,
-      "loss": 0.4021,
+      "loss": 0.4551,
       "step": 550
     },
     {
-      "epoch": 26.08695652173913,
-      "grad_norm": 0.8799042701721191,
+      "epoch": 23.762376237623762,
+      "grad_norm": 0.8126183152198792,
       "learning_rate": 3e-06,
-      "loss": 0.4071,
+      "loss": 0.4542,
       "step": 600
     },
     {
-      "epoch": 28.26086956521739,
-      "grad_norm": 1.0451533794403076,
+      "epoch": 25.742574257425744,
+      "grad_norm": 0.6602805852890015,
       "learning_rate": 3.2500000000000002e-06,
-      "loss": 0.4146,
+      "loss": 0.4535,
       "step": 650
     },
     {
-      "epoch": 30.434782608695652,
-      "grad_norm": 0.7998800277709961,
+      "epoch": 27.722772277227723,
+      "grad_norm": 0.7177000641822815,
       "learning_rate": 3.5e-06,
-      "loss": 0.4023,
+      "loss": 0.4484,
       "step": 700
     },
     {
-      "epoch": 32.608695652173914,
-      "grad_norm": 1.0301792621612549,
+      "epoch": 29.702970297029704,
+      "grad_norm": 0.7912996411323547,
       "learning_rate": 3.7500000000000005e-06,
-      "loss": 0.4045,
+      "loss": 0.4486,
       "step": 750
     },
     {
-      "epoch": 34.78260869565217,
-      "grad_norm": 0.6388185620307922,
+      "epoch": 31.683168316831683,
+      "grad_norm": 0.6251747608184814,
       "learning_rate": 4.000000000000001e-06,
-      "loss": 0.4045,
+      "loss": 0.4463,
       "step": 800
     },
     {
-      "epoch": 36.95652173913044,
-      "grad_norm": 0.9676840901374817,
+      "epoch": 33.663366336633665,
+      "grad_norm": 1.003780722618103,
       "learning_rate": 4.25e-06,
-      "loss": 0.4116,
+      "loss": 0.4474,
       "step": 850
     },
     {
-      "epoch": 39.130434782608695,
-      "grad_norm": 0.8355514407157898,
+      "epoch": 35.64356435643565,
+      "grad_norm": 0.742171585559845,
       "learning_rate": 4.5e-06,
-      "loss": 0.4013,
+      "loss": 0.4434,
       "step": 900
     },
     {
-      "epoch": 41.30434782608695,
-      "grad_norm": 0.7534067630767822,
+      "epoch": 37.62376237623762,
+      "grad_norm": 0.6034647822380066,
       "learning_rate": 4.75e-06,
-      "loss": 0.4019,
+      "loss": 0.4425,
       "step": 950
     },
     {
-      "epoch": 43.47826086956522,
-      "grad_norm": 0.9366117119789124,
+      "epoch": 39.603960396039604,
+      "grad_norm": 0.6990635395050049,
       "learning_rate": 5e-06,
-      "loss": 0.4064,
+      "loss": 0.4381,
       "step": 1000
     },
     {
-      "epoch": 43.47826086956522,
-      "eval_loss": 0.41316837072372437,
-      "eval_runtime": 6.5616,
-      "eval_samples_per_second": 24.232,
-      "eval_steps_per_second": 3.048,
+      "epoch": 39.603960396039604,
+      "eval_loss": 0.41002073884010315,
+      "eval_runtime": 6.7782,
+      "eval_samples_per_second": 26.556,
+      "eval_steps_per_second": 3.393,
       "step": 1000
     },
     {
-      "epoch": 45.65217391304348,
-      "grad_norm": 0.7025179266929626,
+      "epoch": 41.584158415841586,
+      "grad_norm": 0.9420008659362793,
       "learning_rate": 5.2500000000000006e-06,
-      "loss": 0.4061,
+      "loss": 0.4421,
       "step": 1050
     },
     {
-      "epoch": 47.82608695652174,
-      "grad_norm": 0.9240646958351135,
+      "epoch": 43.56435643564357,
+      "grad_norm": 0.7143107652664185,
       "learning_rate": 5.500000000000001e-06,
-      "loss": 0.4061,
+      "loss": 0.4389,
       "step": 1100
     },
     {
-      "epoch": 50.0,
-      "grad_norm": 0.9685391187667847,
+      "epoch": 45.54455445544554,
+      "grad_norm": 0.7934616804122925,
       "learning_rate": 5.75e-06,
-      "loss": 0.4063,
+      "loss": 0.4369,
       "step": 1150
     },
     {
-      "epoch": 52.17391304347826,
-      "grad_norm": 0.8591094017028809,
+      "epoch": 47.524752475247524,
+      "grad_norm": 0.9682563543319702,
       "learning_rate": 6e-06,
-      "loss": 0.4013,
+      "loss": 0.4355,
       "step": 1200
     },
     {
-      "epoch": 54.34782608695652,
-      "grad_norm": 0.9348228573799133,
+      "epoch": 49.504950495049506,
+      "grad_norm": 0.7880070209503174,
       "learning_rate": 6.25e-06,
-      "loss": 0.402,
+      "loss": 0.4359,
       "step": 1250
     },
     {
-      "epoch": 56.52173913043478,
-      "grad_norm": 0.7908409237861633,
+      "epoch": 51.48514851485149,
+      "grad_norm": 0.8443828821182251,
       "learning_rate": 6.5000000000000004e-06,
-      "loss": 0.4046,
+      "loss": 0.4324,
       "step": 1300
     },
     {
-      "epoch": 58.69565217391305,
-      "grad_norm": 1.109017014503479,
+      "epoch": 53.46534653465346,
+      "grad_norm": 0.724743127822876,
       "learning_rate": 6.750000000000001e-06,
-      "loss": 0.3995,
+      "loss": 0.4348,
       "step": 1350
     },
     {
-      "epoch": 60.869565217391305,
-      "grad_norm": 0.8529478311538696,
+      "epoch": 55.445544554455445,
+      "grad_norm": 0.7557791471481323,
       "learning_rate": 7e-06,
-      "loss": 0.4018,
+      "loss": 0.433,
       "step": 1400
     },
     {
-      "epoch": 63.04347826086956,
-      "grad_norm": 0.9947476387023926,
+      "epoch": 57.42574257425743,
+      "grad_norm": 0.7936097383499146,
       "learning_rate": 7.25e-06,
-      "loss": 0.401,
+      "loss": 0.4273,
       "step": 1450
     },
     {
-      "epoch": 65.21739130434783,
-      "grad_norm": 1.2476813793182373,
+      "epoch": 59.40594059405941,
+      "grad_norm": 0.7041624784469604,
       "learning_rate": 7.500000000000001e-06,
-      "loss": 0.4034,
+      "loss": 0.4281,
       "step": 1500
     },
     {
-      "epoch": 65.21739130434783,
-      "eval_loss": 0.41223829984664917,
-      "eval_runtime": 6.5846,
-      "eval_samples_per_second": 24.147,
-      "eval_steps_per_second": 3.037,
+      "epoch": 59.40594059405941,
+      "eval_loss": 0.40218469500541687,
+      "eval_runtime": 7.4309,
+      "eval_samples_per_second": 24.223,
+      "eval_steps_per_second": 3.095,
       "step": 1500
     },
     {
-      "epoch": 67.3913043478261,
-      "grad_norm": 0.8658013343811035,
+      "epoch": 61.386138613861384,
+      "grad_norm": 0.9194286465644836,
       "learning_rate": 7.75e-06,
-      "loss": 0.4007,
+      "loss": 0.4275,
       "step": 1550
     },
     {
-      "epoch": 69.56521739130434,
-      "grad_norm": 0.7932788133621216,
+      "epoch": 63.366336633663366,
+      "grad_norm": 0.7239477634429932,
       "learning_rate": 8.000000000000001e-06,
-      "loss": 0.4054,
+      "loss": 0.4245,
       "step": 1600
     },
     {
-      "epoch": 71.73913043478261,
-      "grad_norm": 2.33347225189209,
+      "epoch": 65.34653465346534,
+      "grad_norm": 0.7235706448554993,
       "learning_rate": 8.25e-06,
-      "loss": 0.4075,
+      "loss": 0.4248,
       "step": 1650
     },
     {
-      "epoch": 73.91304347826087,
-      "grad_norm": 1.0095164775848389,
+      "epoch": 67.32673267326733,
+      "grad_norm": 0.7281640768051147,
       "learning_rate": 8.5e-06,
-      "loss": 0.3982,
+      "loss": 0.425,
       "step": 1700
     },
     {
-      "epoch": 76.08695652173913,
-      "grad_norm": 0.8895902633666992,
+      "epoch": 69.3069306930693,
+      "grad_norm": 1.011294960975647,
       "learning_rate": 8.750000000000001e-06,
-      "loss": 0.4018,
+      "loss": 0.4238,
       "step": 1750
     },
     {
-      "epoch": 78.26086956521739,
-      "grad_norm": 1.2252330780029297,
+      "epoch": 71.2871287128713,
+      "grad_norm": 1.1249107122421265,
       "learning_rate": 9e-06,
-      "loss": 0.4024,
+      "loss": 0.4203,
       "step": 1800
     },
     {
-      "epoch": 80.43478260869566,
-      "grad_norm": 0.7731898427009583,
+      "epoch": 73.26732673267327,
+      "grad_norm": 0.7369679808616638,
       "learning_rate": 9.250000000000001e-06,
-      "loss": 0.3994,
+      "loss": 0.422,
       "step": 1850
     },
     {
-      "epoch": 82.6086956521739,
-      "grad_norm": 1.047607660293579,
+      "epoch": 75.24752475247524,
+      "grad_norm": 0.704367995262146,
       "learning_rate": 9.5e-06,
-      "loss": 0.4013,
+      "loss": 0.4201,
       "step": 1900
     },
     {
-      "epoch": 84.78260869565217,
-      "grad_norm": 0.9284490346908569,
+      "epoch": 77.22772277227723,
+      "grad_norm": 0.8307437896728516,
       "learning_rate": 9.75e-06,
-      "loss": 0.3986,
+      "loss": 0.4203,
       "step": 1950
     },
     {
-      "epoch": 86.95652173913044,
-      "grad_norm": 4.725491046905518,
+      "epoch": 79.20792079207921,
+      "grad_norm": 0.7827133536338806,
       "learning_rate": 1e-05,
-      "loss": 0.3985,
+      "loss": 0.4195,
       "step": 2000
     },
     {
-      "epoch": 86.95652173913044,
-      "eval_loss": 0.4147759974002838,
-      "eval_runtime": 6.5475,
-      "eval_samples_per_second": 24.284,
-      "eval_steps_per_second": 3.055,
+      "epoch": 79.20792079207921,
+      "eval_loss": 0.39878609776496887,
+      "eval_runtime": 8.1579,
+      "eval_samples_per_second": 22.064,
+      "eval_steps_per_second": 2.819,
       "step": 2000
     },
     {
-      "epoch": 89.1304347826087,
-      "grad_norm": 1.1678683757781982,
+      "epoch": 81.18811881188118,
+      "grad_norm": 0.8909313082695007,
       "learning_rate": 9.916666666666668e-06,
-      "loss": 0.4066,
+      "loss": 0.4203,
       "step": 2050
     },
     {
-      "epoch": 91.30434782608695,
-      "grad_norm": 1.988890528678894,
-      "learning_rate": 9.835000000000002e-06,
-      "loss": 0.3997,
+      "epoch": 83.16831683168317,
+      "grad_norm": 0.7147483825683594,
+      "learning_rate": 9.833333333333333e-06,
+      "loss": 0.418,
       "step": 2100
     },
     {
-      "epoch": 93.47826086956522,
-      "grad_norm": 1.0278427600860596,
-      "learning_rate": 9.751666666666667e-06,
-      "loss": 0.3948,
+      "epoch": 85.14851485148515,
+      "grad_norm": 0.6844545006752014,
+      "learning_rate": 9.75e-06,
+      "loss": 0.4167,
       "step": 2150
     },
     {
-      "epoch": 95.65217391304348,
-      "grad_norm": 0.9606735110282898,
-      "learning_rate": 9.67e-06,
-      "loss": 0.42,
+      "epoch": 87.12871287128714,
+      "grad_norm": 0.88512122631073,
+      "learning_rate": 9.666666666666667e-06,
+      "loss": 0.4108,
       "step": 2200
     },
     {
-      "epoch": 97.82608695652173,
-      "grad_norm": 1.0861903429031372,
-      "learning_rate": 9.586666666666667e-06,
-      "loss": 0.398,
+      "epoch": 89.10891089108911,
+      "grad_norm": 0.7097252607345581,
+      "learning_rate": 9.583333333333335e-06,
+      "loss": 0.4135,
       "step": 2250
     },
     {
-      "epoch": 100.0,
-      "grad_norm": 1.4872643947601318,
-      "learning_rate": 9.503333333333334e-06,
-      "loss": 0.4059,
+      "epoch": 91.08910891089108,
+      "grad_norm": 0.8400272727012634,
+      "learning_rate": 9.5e-06,
+      "loss": 0.4147,
       "step": 2300
     },
     {
-      "epoch": 102.17391304347827,
-      "grad_norm": 1.0460981130599976,
-      "learning_rate": 9.42e-06,
-      "loss": 0.4014,
+      "epoch": 93.06930693069307,
+      "grad_norm": 0.7614939212799072,
+      "learning_rate": 9.416666666666667e-06,
+      "loss": 0.4119,
       "step": 2350
     },
     {
-      "epoch": 104.34782608695652,
-      "grad_norm": 1.906476378440857,
-      "learning_rate": 9.336666666666666e-06,
-      "loss": 0.4015,
+      "epoch": 95.04950495049505,
+      "grad_norm": 0.8099732995033264,
+      "learning_rate": 9.333333333333334e-06,
+      "loss": 0.4108,
       "step": 2400
     },
     {
-      "epoch": 106.52173913043478,
-      "grad_norm": 1.3965051174163818,
-      "learning_rate": 9.253333333333333e-06,
-      "loss": 0.4063,
+      "epoch": 97.02970297029702,
+      "grad_norm": 1.0697815418243408,
+      "learning_rate": 9.250000000000001e-06,
+      "loss": 0.4095,
       "step": 2450
     },
     {
-      "epoch": 108.69565217391305,
-      "grad_norm": 1.7410107851028442,
-      "learning_rate": 9.17e-06,
-      "loss": 0.396,
+      "epoch": 99.00990099009901,
+      "grad_norm": 0.7235235571861267,
+      "learning_rate": 9.166666666666666e-06,
+      "loss": 0.4134,
       "step": 2500
     },
     {
-      "epoch": 108.69565217391305,
-      "eval_loss": 0.41233959794044495,
-      "eval_runtime": 6.6031,
-      "eval_samples_per_second": 24.08,
-      "eval_steps_per_second": 3.029,
+      "epoch": 99.00990099009901,
+      "eval_loss": 0.39545077085494995,
+      "eval_runtime": 7.724,
+      "eval_samples_per_second": 23.304,
+      "eval_steps_per_second": 2.978,
       "step": 2500
     }
   ],
   "logging_steps": 50,
   "max_steps": 8000,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 348,
+  "num_train_epochs": 320,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -416,8 +416,8 @@
       "attributes": {}
     }
   },
-  "total_flos": 3.699953465472e+16,
-  "train_batch_size": 64,
+  "total_flos": 2.701326498107371e+16,
+  "train_batch_size": 16,
   "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:f95de2b194fe7f16ec10fbf1b93d83e04c25498d479793b9a5748c1d338bfbd0
+oid sha256:7e97537a0fb344678f807f30d02bd561fb4b5c2162a68338251550dd3ef50d20
 size 5304