Training in progress, step 2500, checkpoint

checkpoint-2500/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:32f3a1545c6b7f8058c4294b66661bbd9e3d4dec0f402db560eba66003a6bb19
+oid sha256:7493ad5637e6ccf113491f9e3ebc087d34378e0d2a945336aceb41119bc02077
 size 577789320

checkpoint-2500/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:02dad2ed8984c6eca1db04c9d3e2c43ae29372f7956bcc4fc7c2e307548955de
+oid sha256:4f370d362c714dd35f5e4e0a87163395f9a19b0a5b2aae1e343376a83b30b2b0
 size 1155772233

checkpoint-2500/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e12cffd706a7a40f72df497f7556e19c62f5295a943d4c2d84107cc3757766d8
+oid sha256:a806455f7b2323ada1257a2ccc10634eadf1cae42d8679a695b446f85f9dcda8
 size 14244

checkpoint-2500/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1cdca242b8590cd159290da20e22d5d5d903dee7a387d8e3f7463cb40559acf6
+oid sha256:f5b56c909bdfd359503d6bcec618fbe35b9e2276a74764820065954757bce0f5
 size 1064

checkpoint-2500/trainer_state.json

@@ -1,7 +1,7 @@
 {
-  "best_metric": 0.42134132981300354,
+  "best_metric": 0.41214704513549805,
   "best_model_checkpoint": "mikhail-panzo/fil_b32_le5_s4000/checkpoint-2500",
-  "epoch": 55.55555555555556,
+  "epoch": 222.22222222222223,
   "eval_steps": 500,
   "global_step": 2500,
   "is_hyper_param_search": false,
@@ -9,400 +9,400 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 1.1111111111111112,
-      "grad_norm": 4.398683071136475,
-      "learning_rate": 2.4500000000000004e-07,
-      "loss": 0.8108,
+      "epoch": 4.444444444444445,
+      "grad_norm": 4.288165092468262,
+      "learning_rate": 2.5000000000000004e-07,
+      "loss": 0.7989,
       "step": 50
     },
     {
-      "epoch": 2.2222222222222223,
-      "grad_norm": 3.726436138153076,
-      "learning_rate": 4.95e-07,
-      "loss": 0.792,
+      "epoch": 8.88888888888889,
+      "grad_norm": 2.4493463039398193,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 0.7701,
       "step": 100
     },
     {
-      "epoch": 3.3333333333333335,
-      "grad_norm": 2.0379676818847656,
-      "learning_rate": 7.450000000000001e-07,
-      "loss": 0.7413,
+      "epoch": 13.333333333333334,
+      "grad_norm": 1.3798907995224,
+      "learning_rate": 7.5e-07,
+      "loss": 0.7374,
       "step": 150
     },
     {
-      "epoch": 4.444444444444445,
-      "grad_norm": 2.0460364818573,
-      "learning_rate": 9.950000000000002e-07,
-      "loss": 0.7342,
+      "epoch": 17.77777777777778,
+      "grad_norm": 2.6173877716064453,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.7101,
       "step": 200
     },
     {
-      "epoch": 5.555555555555555,
-      "grad_norm": 2.803539276123047,
-      "learning_rate": 1.2450000000000002e-06,
-      "loss": 0.7147,
+      "epoch": 22.22222222222222,
+      "grad_norm": 1.3756701946258545,
+      "learning_rate": 1.25e-06,
+      "loss": 0.6887,
       "step": 250
     },
     {
-      "epoch": 6.666666666666667,
-      "grad_norm": 1.9818241596221924,
-      "learning_rate": 1.495e-06,
-      "loss": 0.7072,
+      "epoch": 26.666666666666668,
+      "grad_norm": 1.0456843376159668,
+      "learning_rate": 1.5e-06,
+      "loss": 0.672,
       "step": 300
     },
     {
-      "epoch": 7.777777777777778,
-      "grad_norm": 3.081399917602539,
-      "learning_rate": 1.745e-06,
-      "loss": 0.6843,
+      "epoch": 31.11111111111111,
+      "grad_norm": 0.9518398642539978,
+      "learning_rate": 1.75e-06,
+      "loss": 0.6583,
       "step": 350
     },
     {
-      "epoch": 8.88888888888889,
-      "grad_norm": 2.4465785026550293,
-      "learning_rate": 1.9950000000000004e-06,
-      "loss": 0.657,
+      "epoch": 35.55555555555556,
+      "grad_norm": 1.305978775024414,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.6258,
       "step": 400
     },
     {
-      "epoch": 10.0,
-      "grad_norm": 2.929323434829712,
-      "learning_rate": 2.245e-06,
-      "loss": 0.6597,
+      "epoch": 40.0,
+      "grad_norm": 0.8857623338699341,
+      "learning_rate": 2.25e-06,
+      "loss": 0.5835,
       "step": 450
     },
     {
-      "epoch": 11.11111111111111,
-      "grad_norm": 1.419311285018921,
-      "learning_rate": 2.4950000000000003e-06,
-      "loss": 0.6246,
+      "epoch": 44.44444444444444,
+      "grad_norm": 1.052182912826538,
+      "learning_rate": 2.5e-06,
+      "loss": 0.5539,
       "step": 500
     },
     {
-      "epoch": 11.11111111111111,
-      "eval_loss": 0.5375720262527466,
-      "eval_runtime": 9.2568,
-      "eval_samples_per_second": 17.177,
-      "eval_steps_per_second": 2.161,
+      "epoch": 44.44444444444444,
+      "eval_loss": 0.48496538400650024,
+      "eval_runtime": 9.4574,
+      "eval_samples_per_second": 16.812,
+      "eval_steps_per_second": 2.115,
       "step": 500
     },
     {
-      "epoch": 12.222222222222221,
-      "grad_norm": 2.3438374996185303,
-      "learning_rate": 2.7450000000000004e-06,
-      "loss": 0.6018,
+      "epoch": 48.888888888888886,
+      "grad_norm": 1.249458909034729,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.5387,
       "step": 550
     },
     {
-      "epoch": 13.333333333333334,
-      "grad_norm": 1.7441222667694092,
-      "learning_rate": 2.995e-06,
-      "loss": 0.5547,
+      "epoch": 53.333333333333336,
+      "grad_norm": 0.801243245601654,
+      "learning_rate": 3e-06,
+      "loss": 0.5344,
       "step": 600
     },
     {
-      "epoch": 14.444444444444445,
-      "grad_norm": 1.5592021942138672,
-      "learning_rate": 3.2450000000000003e-06,
-      "loss": 0.5466,
+      "epoch": 57.77777777777778,
+      "grad_norm": 1.1922223567962646,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.5259,
       "step": 650
     },
     {
-      "epoch": 15.555555555555555,
-      "grad_norm": 1.305444359779358,
-      "learning_rate": 3.495e-06,
-      "loss": 0.535,
+      "epoch": 62.22222222222222,
+      "grad_norm": 0.9428519606590271,
+      "learning_rate": 3.5e-06,
+      "loss": 0.5179,
       "step": 700
     },
     {
-      "epoch": 16.666666666666668,
-      "grad_norm": 2.2067477703094482,
-      "learning_rate": 3.745e-06,
-      "loss": 0.5343,
+      "epoch": 66.66666666666667,
+      "grad_norm": 0.755260705947876,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.5082,
       "step": 750
     },
     {
-      "epoch": 17.77777777777778,
-      "grad_norm": 1.667480707168579,
-      "learning_rate": 3.995000000000001e-06,
-      "loss": 0.5296,
+      "epoch": 71.11111111111111,
+      "grad_norm": 1.8587590456008911,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.5096,
       "step": 800
     },
     {
-      "epoch": 18.88888888888889,
-      "grad_norm": 1.5465797185897827,
-      "learning_rate": 4.245e-06,
-      "loss": 0.5198,
+      "epoch": 75.55555555555556,
+      "grad_norm": 0.6854829788208008,
+      "learning_rate": 4.25e-06,
+      "loss": 0.4952,
       "step": 850
     },
     {
-      "epoch": 20.0,
-      "grad_norm": 2.760589361190796,
-      "learning_rate": 4.495e-06,
-      "loss": 0.5193,
+      "epoch": 80.0,
+      "grad_norm": 0.9088854789733887,
+      "learning_rate": 4.5e-06,
+      "loss": 0.4941,
       "step": 900
     },
     {
-      "epoch": 21.11111111111111,
-      "grad_norm": 1.4570050239562988,
-      "learning_rate": 4.745e-06,
-      "loss": 0.5124,
+      "epoch": 84.44444444444444,
+      "grad_norm": 0.8932746648788452,
+      "learning_rate": 4.75e-06,
+      "loss": 0.4936,
       "step": 950
     },
     {
-      "epoch": 22.22222222222222,
-      "grad_norm": 2.1914265155792236,
-      "learning_rate": 4.9950000000000005e-06,
-      "loss": 0.5034,
+      "epoch": 88.88888888888889,
+      "grad_norm": 0.9161014556884766,
+      "learning_rate": 5e-06,
+      "loss": 0.4867,
       "step": 1000
     },
     {
-      "epoch": 22.22222222222222,
-      "eval_loss": 0.45661377906799316,
-      "eval_runtime": 9.2158,
-      "eval_samples_per_second": 17.253,
-      "eval_steps_per_second": 2.17,
+      "epoch": 88.88888888888889,
+      "eval_loss": 0.4438903331756592,
+      "eval_runtime": 9.2831,
+      "eval_samples_per_second": 17.128,
+      "eval_steps_per_second": 2.154,
       "step": 1000
     },
     {
-      "epoch": 23.333333333333332,
-      "grad_norm": 1.941751480102539,
-      "learning_rate": 5.245e-06,
-      "loss": 0.5046,
+      "epoch": 93.33333333333333,
+      "grad_norm": 0.6858564615249634,
+      "learning_rate": 5.2500000000000006e-06,
+      "loss": 0.486,
       "step": 1050
     },
     {
-      "epoch": 24.444444444444443,
-      "grad_norm": 1.9293475151062012,
-      "learning_rate": 5.495000000000001e-06,
-      "loss": 0.4948,
+      "epoch": 97.77777777777777,
+      "grad_norm": 1.2040996551513672,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 0.4789,
       "step": 1100
     },
     {
-      "epoch": 25.555555555555557,
-      "grad_norm": 1.3278849124908447,
-      "learning_rate": 5.745000000000001e-06,
-      "loss": 0.4931,
+      "epoch": 102.22222222222223,
+      "grad_norm": 1.0067086219787598,
+      "learning_rate": 5.75e-06,
+      "loss": 0.477,
       "step": 1150
     },
     {
-      "epoch": 26.666666666666668,
-      "grad_norm": 1.1657841205596924,
-      "learning_rate": 5.995000000000001e-06,
-      "loss": 0.496,
+      "epoch": 106.66666666666667,
+      "grad_norm": 1.4230589866638184,
+      "learning_rate": 6e-06,
+      "loss": 0.473,
       "step": 1200
     },
     {
-      "epoch": 27.77777777777778,
-      "grad_norm": 1.8730145692825317,
-      "learning_rate": 6.245000000000001e-06,
-      "loss": 0.4939,
+      "epoch": 111.11111111111111,
+      "grad_norm": 0.916904091835022,
+      "learning_rate": 6.25e-06,
+      "loss": 0.4698,
       "step": 1250
     },
     {
-      "epoch": 28.88888888888889,
-      "grad_norm": 2.1156740188598633,
-      "learning_rate": 6.4950000000000005e-06,
-      "loss": 0.4908,
+      "epoch": 115.55555555555556,
+      "grad_norm": 0.9870907664299011,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 0.4693,
       "step": 1300
     },
     {
-      "epoch": 30.0,
-      "grad_norm": 1.799806833267212,
-      "learning_rate": 6.745000000000001e-06,
-      "loss": 0.4875,
+      "epoch": 120.0,
+      "grad_norm": 1.0495694875717163,
+      "learning_rate": 6.750000000000001e-06,
+      "loss": 0.4651,
       "step": 1350
     },
     {
-      "epoch": 31.11111111111111,
-      "grad_norm": 1.762253761291504,
-      "learning_rate": 6.995000000000001e-06,
-      "loss": 0.4836,
+      "epoch": 124.44444444444444,
+      "grad_norm": 1.6615877151489258,
+      "learning_rate": 7e-06,
+      "loss": 0.4648,
       "step": 1400
     },
     {
-      "epoch": 32.22222222222222,
-      "grad_norm": 1.305191993713379,
-      "learning_rate": 7.245000000000001e-06,
-      "loss": 0.4811,
+      "epoch": 128.88888888888889,
+      "grad_norm": 0.89169842004776,
+      "learning_rate": 7.25e-06,
+      "loss": 0.4643,
       "step": 1450
     },
     {
-      "epoch": 33.333333333333336,
-      "grad_norm": 1.5870022773742676,
-      "learning_rate": 7.495000000000001e-06,
-      "loss": 0.4807,
+      "epoch": 133.33333333333334,
+      "grad_norm": 0.8503329157829285,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.4595,
       "step": 1500
     },
     {
-      "epoch": 33.333333333333336,
-      "eval_loss": 0.43601250648498535,
-      "eval_runtime": 9.5435,
-      "eval_samples_per_second": 16.66,
-      "eval_steps_per_second": 2.096,
+      "epoch": 133.33333333333334,
+      "eval_loss": 0.4245499074459076,
+      "eval_runtime": 8.4972,
+      "eval_samples_per_second": 18.712,
+      "eval_steps_per_second": 2.354,
       "step": 1500
     },
     {
-      "epoch": 34.44444444444444,
-      "grad_norm": 2.632877826690674,
-      "learning_rate": 7.745e-06,
-      "loss": 0.4782,
+      "epoch": 137.77777777777777,
+      "grad_norm": 0.9252403974533081,
+      "learning_rate": 7.75e-06,
+      "loss": 0.4617,
       "step": 1550
     },
     {
-      "epoch": 35.55555555555556,
-      "grad_norm": 1.3705276250839233,
-      "learning_rate": 7.995e-06,
-      "loss": 0.48,
+      "epoch": 142.22222222222223,
+      "grad_norm": 0.8366873860359192,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.4557,
       "step": 1600
     },
     {
-      "epoch": 36.666666666666664,
-      "grad_norm": 1.9232587814331055,
-      "learning_rate": 8.245000000000002e-06,
-      "loss": 0.47,
+      "epoch": 146.66666666666666,
+      "grad_norm": 0.7907727360725403,
+      "learning_rate": 8.25e-06,
+      "loss": 0.4565,
       "step": 1650
     },
     {
-      "epoch": 37.77777777777778,
-      "grad_norm": 1.3885018825531006,
-      "learning_rate": 8.495e-06,
-      "loss": 0.4711,
+      "epoch": 151.11111111111111,
+      "grad_norm": 0.8415725231170654,
+      "learning_rate": 8.5e-06,
+      "loss": 0.4544,
       "step": 1700
     },
     {
-      "epoch": 38.888888888888886,
-      "grad_norm": 2.4857094287872314,
-      "learning_rate": 8.745000000000002e-06,
-      "loss": 0.4721,
+      "epoch": 155.55555555555554,
+      "grad_norm": 0.8423948884010315,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 0.4528,
       "step": 1750
     },
     {
-      "epoch": 40.0,
-      "grad_norm": 1.7662663459777832,
-      "learning_rate": 8.995000000000001e-06,
-      "loss": 0.4656,
+      "epoch": 160.0,
+      "grad_norm": 0.8236134648323059,
+      "learning_rate": 9e-06,
+      "loss": 0.4555,
       "step": 1800
     },
     {
-      "epoch": 41.111111111111114,
-      "grad_norm": 1.906548261642456,
-      "learning_rate": 9.245e-06,
-      "loss": 0.463,
+      "epoch": 164.44444444444446,
+      "grad_norm": 0.8070024251937866,
+      "learning_rate": 9.250000000000001e-06,
+      "loss": 0.4474,
       "step": 1850
     },
     {
-      "epoch": 42.22222222222222,
-      "grad_norm": 1.7204883098602295,
-      "learning_rate": 9.495000000000001e-06,
-      "loss": 0.4592,
+      "epoch": 168.88888888888889,
+      "grad_norm": 1.041704535484314,
+      "learning_rate": 9.5e-06,
+      "loss": 0.4478,
       "step": 1900
     },
     {
-      "epoch": 43.333333333333336,
-      "grad_norm": 1.3313167095184326,
-      "learning_rate": 9.745e-06,
-      "loss": 0.4581,
+      "epoch": 173.33333333333334,
+      "grad_norm": 1.021083116531372,
+      "learning_rate": 9.75e-06,
+      "loss": 0.4433,
       "step": 1950
     },
     {
-      "epoch": 44.44444444444444,
-      "grad_norm": 1.94256591796875,
-      "learning_rate": 9.995000000000002e-06,
-      "loss": 0.4582,
+      "epoch": 177.77777777777777,
+      "grad_norm": 1.097479224205017,
+      "learning_rate": 1e-05,
+      "loss": 0.4395,
       "step": 2000
     },
     {
-      "epoch": 44.44444444444444,
-      "eval_loss": 0.4244868755340576,
-      "eval_runtime": 9.1316,
-      "eval_samples_per_second": 17.412,
-      "eval_steps_per_second": 2.19,
+      "epoch": 177.77777777777777,
+      "eval_loss": 0.41554170846939087,
+      "eval_runtime": 9.3691,
+      "eval_samples_per_second": 16.971,
+      "eval_steps_per_second": 2.135,
       "step": 2000
     },
     {
-      "epoch": 45.55555555555556,
-      "grad_norm": 2.019158124923706,
-      "learning_rate": 9.755e-06,
-      "loss": 0.4573,
+      "epoch": 182.22222222222223,
+      "grad_norm": 1.0658737421035767,
+      "learning_rate": 9.75e-06,
+      "loss": 0.4403,
       "step": 2050
     },
     {
-      "epoch": 46.666666666666664,
-      "grad_norm": 1.7636100053787231,
-      "learning_rate": 9.505000000000001e-06,
-      "loss": 0.4642,
+      "epoch": 186.66666666666666,
+      "grad_norm": 0.7741203308105469,
+      "learning_rate": 9.5e-06,
+      "loss": 0.4371,
       "step": 2100
     },
     {
-      "epoch": 47.77777777777778,
-      "grad_norm": 2.0878355503082275,
-      "learning_rate": 9.260000000000001e-06,
-      "loss": 0.454,
+      "epoch": 191.11111111111111,
+      "grad_norm": 0.8409827947616577,
+      "learning_rate": 9.250000000000001e-06,
+      "loss": 0.4383,
       "step": 2150
     },
     {
-      "epoch": 48.888888888888886,
-      "grad_norm": 1.5207270383834839,
-      "learning_rate": 9.01e-06,
-      "loss": 0.4573,
+      "epoch": 195.55555555555554,
+      "grad_norm": 0.7340779900550842,
+      "learning_rate": 9e-06,
+      "loss": 0.4367,
       "step": 2200
     },
     {
-      "epoch": 50.0,
-      "grad_norm": 2.592440605163574,
-      "learning_rate": 8.76e-06,
-      "loss": 0.4557,
+      "epoch": 200.0,
+      "grad_norm": 1.0130568742752075,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 0.4364,
       "step": 2250
     },
     {
-      "epoch": 51.111111111111114,
-      "grad_norm": 4.970732688903809,
-      "learning_rate": 8.51e-06,
-      "loss": 0.4565,
+      "epoch": 204.44444444444446,
+      "grad_norm": 0.7156779170036316,
+      "learning_rate": 8.5e-06,
+      "loss": 0.4337,
       "step": 2300
     },
     {
-      "epoch": 52.22222222222222,
-      "grad_norm": 1.2565674781799316,
-      "learning_rate": 8.26e-06,
-      "loss": 0.4492,
+      "epoch": 208.88888888888889,
+      "grad_norm": 1.1859618425369263,
+      "learning_rate": 8.25e-06,
+      "loss": 0.4342,
       "step": 2350
     },
     {
-      "epoch": 53.333333333333336,
-      "grad_norm": 1.402048110961914,
-      "learning_rate": 8.010000000000001e-06,
-      "loss": 0.4477,
+      "epoch": 213.33333333333334,
+      "grad_norm": 0.6573166847229004,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.432,
       "step": 2400
     },
     {
-      "epoch": 54.44444444444444,
-      "grad_norm": 2.7496023178100586,
-      "learning_rate": 7.76e-06,
-      "loss": 0.4484,
+      "epoch": 217.77777777777777,
+      "grad_norm": 1.0668668746948242,
+      "learning_rate": 7.75e-06,
+      "loss": 0.4338,
       "step": 2450
     },
     {
-      "epoch": 55.55555555555556,
-      "grad_norm": 2.6707046031951904,
-      "learning_rate": 7.510000000000001e-06,
-      "loss": 0.4541,
+      "epoch": 222.22222222222223,
+      "grad_norm": 0.8622764945030212,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.4327,
       "step": 2500
     },
     {
-      "epoch": 55.55555555555556,
-      "eval_loss": 0.42134132981300354,
-      "eval_runtime": 9.5109,
-      "eval_samples_per_second": 16.718,
-      "eval_steps_per_second": 2.103,
+      "epoch": 222.22222222222223,
+      "eval_loss": 0.41214704513549805,
+      "eval_runtime": 9.4587,
+      "eval_samples_per_second": 16.81,
+      "eval_steps_per_second": 2.114,
       "step": 2500
     }
   ],
   "logging_steps": 50,
   "max_steps": 4000,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 89,
+  "num_train_epochs": 364,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -416,7 +416,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.7199264863629728e+16,
+  "total_flos": 6.872445306432288e+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 sha256:05b7ecdded947540b9670cedf7cfe87afc7199b237efb642b915ffbb505d51dc
+oid sha256:2de69a241d012253ce4b1526ecd0f9fd0f7dfbedfea74bfff08144d42d9f523a
 size 5304