itsLeen/vit-large-ai-or-not

README.md

@@ -18,8 +18,8 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google/vit-large-patch16-224](https://huggingface.co/google/vit-large-patch16-224) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1641
-- Accuracy: 0.9761
+- Loss: 0.1282
+- Accuracy: 0.9654
 
 ## Model description
 

all_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 2.0,
-    "total_flos": 9.690147678529511e+18,
-    "train_loss": 0.16445749211268226,
-    "train_runtime": 2512.7977,
-    "train_samples_per_second": 14.078,
-    "train_steps_per_second": 0.88
+    "epoch": 4.0,
+    "total_flos": 1.668149581703322e+19,
+    "train_loss": 0.03976387239464436,
+    "train_runtime": 7938.8524,
+    "train_samples_per_second": 7.504,
+    "train_steps_per_second": 0.938
 }

runs/Oct23_10-04-38_b03069d3f47e/events.out.tfevents.1729686694.b03069d3f47e.444.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:495c0d0b7d13d4fa9059fa3dda855f76ea87de3f80faf12c9b3c8f153a5dcb11
+size 411

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 2.0,
-    "total_flos": 9.690147678529511e+18,
-    "train_loss": 0.16445749211268226,
-    "train_runtime": 2512.7977,
-    "train_samples_per_second": 14.078,
-    "train_steps_per_second": 0.88
+    "epoch": 4.0,
+    "total_flos": 1.668149581703322e+19,
+    "train_loss": 0.03976387239464436,
+    "train_runtime": 7938.8524,
+    "train_samples_per_second": 7.504,
+    "train_steps_per_second": 0.938
 }

trainer_state.json

@@ -1,203 +1,619 @@
 {
-  "best_metric": 0.10391458868980408,
-  "best_model_checkpoint": "vit-large-ai-or-not/checkpoint-2000",
-  "epoch": 2.0,
+  "best_metric": 0.12822793424129486,
+  "best_model_checkpoint": "vit-large-ai-or-not/checkpoint-1600",
+  "epoch": 4.0,
   "eval_steps": 200,
-  "global_step": 2212,
+  "global_step": 7448,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.18083182640144665,
-      "grad_norm": 1.507786750793457,
-      "learning_rate": 0.00018191681735985535,
-      "loss": 0.3925,
+      "epoch": 0.10741138560687433,
+      "grad_norm": 0.16569846868515015,
+      "learning_rate": 9.731471535982815e-06,
+      "loss": 0.1089,
       "step": 200
     },
     {
-      "epoch": 0.18083182640144665,
-      "eval_accuracy": 0.878625134264232,
-      "eval_loss": 0.40452271699905396,
-      "eval_runtime": 14.0535,
-      "eval_samples_per_second": 66.247,
-      "eval_steps_per_second": 8.325,
+      "epoch": 0.10741138560687433,
+      "eval_accuracy": 0.9497851772287863,
+      "eval_loss": 0.17561669647693634,
+      "eval_runtime": 60.3674,
+      "eval_samples_per_second": 61.689,
+      "eval_steps_per_second": 7.719,
       "step": 200
     },
     {
-      "epoch": 0.3616636528028933,
-      "grad_norm": 3.166074514389038,
-      "learning_rate": 0.00016383363471971068,
-      "loss": 0.2803,
+      "epoch": 0.21482277121374865,
+      "grad_norm": 0.006833943538367748,
+      "learning_rate": 9.462943071965628e-06,
+      "loss": 0.041,
       "step": 400
     },
     {
-      "epoch": 0.3616636528028933,
-      "eval_accuracy": 0.9044038668098818,
-      "eval_loss": 0.2385927140712738,
-      "eval_runtime": 13.6284,
-      "eval_samples_per_second": 68.313,
-      "eval_steps_per_second": 8.585,
+      "epoch": 0.21482277121374865,
+      "eval_accuracy": 0.9503222341568206,
+      "eval_loss": 0.21512643992900848,
+      "eval_runtime": 62.0849,
+      "eval_samples_per_second": 59.982,
+      "eval_steps_per_second": 7.506,
       "step": 400
     },
     {
-      "epoch": 0.5424954792043399,
-      "grad_norm": 3.2524709701538086,
-      "learning_rate": 0.00014575045207956602,
-      "loss": 0.2235,
+      "epoch": 0.322234156820623,
+      "grad_norm": 0.010011733509600163,
+      "learning_rate": 9.194414607948443e-06,
+      "loss": 0.0566,
       "step": 600
     },
     {
-      "epoch": 0.5424954792043399,
-      "eval_accuracy": 0.9172932330827067,
-      "eval_loss": 0.1893242746591568,
-      "eval_runtime": 13.8787,
-      "eval_samples_per_second": 67.081,
-      "eval_steps_per_second": 8.43,
+      "epoch": 0.322234156820623,
+      "eval_accuracy": 0.9511278195488722,
+      "eval_loss": 0.2124553769826889,
+      "eval_runtime": 61.5803,
+      "eval_samples_per_second": 60.474,
+      "eval_steps_per_second": 7.567,
       "step": 600
     },
     {
-      "epoch": 0.7233273056057866,
-      "grad_norm": 0.6501625776290894,
-      "learning_rate": 0.00012766726943942136,
-      "loss": 0.217,
+      "epoch": 0.4296455424274973,
+      "grad_norm": 22.38845443725586,
+      "learning_rate": 8.927228786251343e-06,
+      "loss": 0.1028,
       "step": 800
     },
     {
-      "epoch": 0.7233273056057866,
-      "eval_accuracy": 0.9398496240601504,
-      "eval_loss": 0.15967358648777008,
-      "eval_runtime": 15.6768,
-      "eval_samples_per_second": 59.387,
-      "eval_steps_per_second": 7.463,
+      "epoch": 0.4296455424274973,
+      "eval_accuracy": 0.9449516648764769,
+      "eval_loss": 0.20836062729358673,
+      "eval_runtime": 61.4329,
+      "eval_samples_per_second": 60.619,
+      "eval_steps_per_second": 7.586,
       "step": 800
     },
     {
-      "epoch": 0.9041591320072333,
-      "grad_norm": 0.8514215350151062,
-      "learning_rate": 0.00010958408679927667,
-      "loss": 0.1865,
+      "epoch": 0.5370569280343717,
+      "grad_norm": 0.2941912114620209,
+      "learning_rate": 8.658700322234156e-06,
+      "loss": 0.1722,
       "step": 1000
     },
     {
-      "epoch": 0.9041591320072333,
-      "eval_accuracy": 0.9419978517722879,
-      "eval_loss": 0.14128026366233826,
-      "eval_runtime": 14.4065,
-      "eval_samples_per_second": 64.624,
-      "eval_steps_per_second": 8.121,
+      "epoch": 0.5370569280343717,
+      "eval_accuracy": 0.9556928034371643,
+      "eval_loss": 0.16581448912620544,
+      "eval_runtime": 61.8304,
+      "eval_samples_per_second": 60.229,
+      "eval_steps_per_second": 7.537,
       "step": 1000
     },
     {
-      "epoch": 1.0849909584086799,
-      "grad_norm": 0.2293129414319992,
-      "learning_rate": 9.150090415913202e-05,
-      "loss": 0.1309,
+      "epoch": 0.644468313641246,
+      "grad_norm": 2.7486720085144043,
+      "learning_rate": 8.390171858216972e-06,
+      "loss": 0.1486,
       "step": 1200
     },
     {
-      "epoch": 1.0849909584086799,
-      "eval_accuracy": 0.9516648764769066,
-      "eval_loss": 0.1473875343799591,
-      "eval_runtime": 14.0786,
-      "eval_samples_per_second": 66.129,
-      "eval_steps_per_second": 8.31,
+      "epoch": 0.644468313641246,
+      "eval_accuracy": 0.959452201933405,
+      "eval_loss": 0.13117314875125885,
+      "eval_runtime": 61.611,
+      "eval_samples_per_second": 60.444,
+      "eval_steps_per_second": 7.564,
       "step": 1200
     },
     {
-      "epoch": 1.2658227848101267,
-      "grad_norm": 4.756102561950684,
-      "learning_rate": 7.341772151898734e-05,
-      "loss": 0.1008,
+      "epoch": 0.7518796992481203,
+      "grad_norm": 7.503540515899658,
+      "learning_rate": 8.121643394199787e-06,
+      "loss": 0.1446,
       "step": 1400
     },
     {
-      "epoch": 1.2658227848101267,
-      "eval_accuracy": 0.9419978517722879,
-      "eval_loss": 0.19141799211502075,
-      "eval_runtime": 13.9995,
-      "eval_samples_per_second": 66.502,
-      "eval_steps_per_second": 8.357,
+      "epoch": 0.7518796992481203,
+      "eval_accuracy": 0.9564983888292159,
+      "eval_loss": 0.16343578696250916,
+      "eval_runtime": 61.6178,
+      "eval_samples_per_second": 60.437,
+      "eval_steps_per_second": 7.563,
       "step": 1400
     },
     {
-      "epoch": 1.4466546112115732,
-      "grad_norm": 0.034902941435575485,
-      "learning_rate": 5.533453887884268e-05,
-      "loss": 0.0793,
+      "epoch": 0.8592910848549946,
+      "grad_norm": 15.931679725646973,
+      "learning_rate": 7.8531149301826e-06,
+      "loss": 0.1281,
       "step": 1600
     },
     {
-      "epoch": 1.4466546112115732,
-      "eval_accuracy": 0.9441460794844253,
-      "eval_loss": 0.15568311512470245,
-      "eval_runtime": 14.1017,
-      "eval_samples_per_second": 66.021,
-      "eval_steps_per_second": 8.297,
+      "epoch": 0.8592910848549946,
+      "eval_accuracy": 0.965359828141783,
+      "eval_loss": 0.12822793424129486,
+      "eval_runtime": 62.167,
+      "eval_samples_per_second": 59.903,
+      "eval_steps_per_second": 7.496,
       "step": 1600
     },
     {
-      "epoch": 1.6274864376130198,
-      "grad_norm": 1.2290639877319336,
-      "learning_rate": 3.725135623869802e-05,
-      "loss": 0.0804,
+      "epoch": 0.966702470461869,
+      "grad_norm": 38.17241668701172,
+      "learning_rate": 7.5845864661654145e-06,
+      "loss": 0.1584,
       "step": 1800
     },
     {
-      "epoch": 1.6274864376130198,
-      "eval_accuracy": 0.9312567132116004,
-      "eval_loss": 0.23011024296283722,
-      "eval_runtime": 13.8445,
-      "eval_samples_per_second": 67.247,
-      "eval_steps_per_second": 8.451,
+      "epoch": 0.966702470461869,
+      "eval_accuracy": 0.966702470461869,
+      "eval_loss": 0.12948955595493317,
+      "eval_runtime": 61.1637,
+      "eval_samples_per_second": 60.886,
+      "eval_steps_per_second": 7.619,
       "step": 1800
     },
     {
-      "epoch": 1.8083182640144666,
-      "grad_norm": 0.08316856622695923,
-      "learning_rate": 1.9168173598553345e-05,
-      "loss": 0.0814,
+      "epoch": 1.0741138560687433,
+      "grad_norm": 0.005669532343745232,
+      "learning_rate": 7.316058002148228e-06,
+      "loss": 0.0549,
       "step": 2000
     },
     {
-      "epoch": 1.8083182640144666,
-      "eval_accuracy": 0.958109559613319,
-      "eval_loss": 0.10391458868980408,
-      "eval_runtime": 14.1201,
-      "eval_samples_per_second": 65.935,
-      "eval_steps_per_second": 8.286,
+      "epoch": 1.0741138560687433,
+      "eval_accuracy": 0.9669709989258861,
+      "eval_loss": 0.16132378578186035,
+      "eval_runtime": 61.0423,
+      "eval_samples_per_second": 61.007,
+      "eval_steps_per_second": 7.634,
       "step": 2000
     },
     {
-      "epoch": 1.9891500904159132,
-      "grad_norm": 0.11144193261861801,
-      "learning_rate": 1.08499095840868e-06,
-      "loss": 0.0446,
+      "epoch": 1.1815252416756177,
+      "grad_norm": 0.003028369741514325,
+      "learning_rate": 7.047529538131043e-06,
+      "loss": 0.0373,
       "step": 2200
     },
     {
-      "epoch": 1.9891500904159132,
-      "eval_accuracy": 0.9634801288936627,
-      "eval_loss": 0.1123916506767273,
-      "eval_runtime": 14.6416,
-      "eval_samples_per_second": 63.586,
-      "eval_steps_per_second": 7.991,
+      "epoch": 1.1815252416756177,
+      "eval_accuracy": 0.9723415682062299,
+      "eval_loss": 0.1343841254711151,
+      "eval_runtime": 62.4644,
+      "eval_samples_per_second": 59.618,
+      "eval_steps_per_second": 7.46,
       "step": 2200
     },
     {
-      "epoch": 2.0,
-      "step": 2212,
-      "total_flos": 9.690147678529511e+18,
-      "train_loss": 0.16445749211268226,
-      "train_runtime": 2512.7977,
-      "train_samples_per_second": 14.078,
-      "train_steps_per_second": 0.88
+      "epoch": 1.2889366272824918,
+      "grad_norm": 0.009722361341118813,
+      "learning_rate": 6.779001074113857e-06,
+      "loss": 0.0293,
+      "step": 2400
+    },
+    {
+      "epoch": 1.2889366272824918,
+      "eval_accuracy": 0.9699248120300752,
+      "eval_loss": 0.15843084454536438,
+      "eval_runtime": 61.4796,
+      "eval_samples_per_second": 60.573,
+      "eval_steps_per_second": 7.58,
+      "step": 2400
+    },
+    {
+      "epoch": 1.3963480128893662,
+      "grad_norm": 0.01564161479473114,
+      "learning_rate": 6.510472610096671e-06,
+      "loss": 0.0251,
+      "step": 2600
+    },
+    {
+      "epoch": 1.3963480128893662,
+      "eval_accuracy": 0.9656283566058003,
+      "eval_loss": 0.17043128609657288,
+      "eval_runtime": 61.4294,
+      "eval_samples_per_second": 60.622,
+      "eval_steps_per_second": 7.586,
+      "step": 2600
+    },
+    {
+      "epoch": 1.5037593984962405,
+      "grad_norm": 0.004944147542119026,
+      "learning_rate": 6.241944146079485e-06,
+      "loss": 0.0249,
+      "step": 2800
+    },
+    {
+      "epoch": 1.5037593984962405,
+      "eval_accuracy": 0.9699248120300752,
+      "eval_loss": 0.1586124300956726,
+      "eval_runtime": 63.1105,
+      "eval_samples_per_second": 59.008,
+      "eval_steps_per_second": 7.384,
+      "step": 2800
+    },
+    {
+      "epoch": 1.6111707841031149,
+      "grad_norm": 0.004947973415255547,
+      "learning_rate": 5.9734156820622995e-06,
+      "loss": 0.0383,
+      "step": 3000
+    },
+    {
+      "epoch": 1.6111707841031149,
+      "eval_accuracy": 0.9715359828141783,
+      "eval_loss": 0.1466864049434662,
+      "eval_runtime": 61.3469,
+      "eval_samples_per_second": 60.704,
+      "eval_steps_per_second": 7.596,
+      "step": 3000
+    },
+    {
+      "epoch": 1.7185821697099892,
+      "grad_norm": 0.00027971353847533464,
+      "learning_rate": 5.704887218045113e-06,
+      "loss": 0.0213,
+      "step": 3200
+    },
+    {
+      "epoch": 1.7185821697099892,
+      "eval_accuracy": 0.9734156820622986,
+      "eval_loss": 0.15455451607704163,
+      "eval_runtime": 61.872,
+      "eval_samples_per_second": 60.189,
+      "eval_steps_per_second": 7.532,
+      "step": 3200
+    },
+    {
+      "epoch": 1.8259935553168636,
+      "grad_norm": 0.0011353939771652222,
+      "learning_rate": 5.436358754027927e-06,
+      "loss": 0.0544,
+      "step": 3400
+    },
+    {
+      "epoch": 1.8259935553168636,
+      "eval_accuracy": 0.9685821697099892,
+      "eval_loss": 0.16710036993026733,
+      "eval_runtime": 62.8836,
+      "eval_samples_per_second": 59.22,
+      "eval_steps_per_second": 7.411,
+      "step": 3400
+    },
+    {
+      "epoch": 1.933404940923738,
+      "grad_norm": 0.021713044494390488,
+      "learning_rate": 5.1678302900107415e-06,
+      "loss": 0.0401,
+      "step": 3600
+    },
+    {
+      "epoch": 1.933404940923738,
+      "eval_accuracy": 0.9656283566058003,
+      "eval_loss": 0.18700723350048065,
+      "eval_runtime": 62.2418,
+      "eval_samples_per_second": 59.831,
+      "eval_steps_per_second": 7.487,
+      "step": 3600
+    },
+    {
+      "epoch": 2.0408163265306123,
+      "grad_norm": 0.010888410732150078,
+      "learning_rate": 4.900644468313642e-06,
+      "loss": 0.0288,
+      "step": 3800
+    },
+    {
+      "epoch": 2.0408163265306123,
+      "eval_accuracy": 0.9599892588614393,
+      "eval_loss": 0.19811120629310608,
+      "eval_runtime": 62.2807,
+      "eval_samples_per_second": 59.794,
+      "eval_steps_per_second": 7.482,
+      "step": 3800
+    },
+    {
+      "epoch": 2.1482277121374866,
+      "grad_norm": 0.0020080420654267073,
+      "learning_rate": 4.632116004296456e-06,
+      "loss": 0.0078,
+      "step": 4000
+    },
+    {
+      "epoch": 2.1482277121374866,
+      "eval_accuracy": 0.9747583243823845,
+      "eval_loss": 0.14224101603031158,
+      "eval_runtime": 61.9174,
+      "eval_samples_per_second": 60.145,
+      "eval_steps_per_second": 7.526,
+      "step": 4000
+    },
+    {
+      "epoch": 2.255639097744361,
+      "grad_norm": 0.0014250120148062706,
+      "learning_rate": 4.36358754027927e-06,
+      "loss": 0.0037,
+      "step": 4200
+    },
+    {
+      "epoch": 2.255639097744361,
+      "eval_accuracy": 0.9704618689581096,
+      "eval_loss": 0.17747652530670166,
+      "eval_runtime": 62.7966,
+      "eval_samples_per_second": 59.303,
+      "eval_steps_per_second": 7.421,
+      "step": 4200
+    },
+    {
+      "epoch": 2.3630504833512354,
+      "grad_norm": 0.0010213665664196014,
+      "learning_rate": 4.095059076262084e-06,
+      "loss": 0.0035,
+      "step": 4400
+    },
+    {
+      "epoch": 2.3630504833512354,
+      "eval_accuracy": 0.9704618689581096,
+      "eval_loss": 0.18454258143901825,
+      "eval_runtime": 62.428,
+      "eval_samples_per_second": 59.653,
+      "eval_steps_per_second": 7.465,
+      "step": 4400
+    },
+    {
+      "epoch": 2.4704618689581097,
+      "grad_norm": 0.0008432798786088824,
+      "learning_rate": 3.826530612244898e-06,
+      "loss": 0.0043,
+      "step": 4600
+    },
+    {
+      "epoch": 2.4704618689581097,
+      "eval_accuracy": 0.9709989258861439,
+      "eval_loss": 0.20010964572429657,
+      "eval_runtime": 61.7124,
+      "eval_samples_per_second": 60.344,
+      "eval_steps_per_second": 7.551,
+      "step": 4600
+    },
+    {
+      "epoch": 2.5778732545649836,
+      "grad_norm": 0.0009041269076988101,
+      "learning_rate": 3.5580021482277123e-06,
+      "loss": 0.0049,
+      "step": 4800
+    },
+    {
+      "epoch": 2.5778732545649836,
+      "eval_accuracy": 0.9688506981740065,
+      "eval_loss": 0.21447789669036865,
+      "eval_runtime": 63.1489,
+      "eval_samples_per_second": 58.972,
+      "eval_steps_per_second": 7.379,
+      "step": 4800
+    },
+    {
+      "epoch": 2.685284640171858,
+      "grad_norm": 0.0005904084537178278,
+      "learning_rate": 3.289473684210527e-06,
+      "loss": 0.01,
+      "step": 5000
+    },
+    {
+      "epoch": 2.685284640171858,
+      "eval_accuracy": 0.9750268528464017,
+      "eval_loss": 0.1444501280784607,
+      "eval_runtime": 61.7502,
+      "eval_samples_per_second": 60.308,
+      "eval_steps_per_second": 7.547,
+      "step": 5000
+    },
+    {
+      "epoch": 2.7926960257787323,
+      "grad_norm": 0.002195934997871518,
+      "learning_rate": 3.020945220193341e-06,
+      "loss": 0.0039,
+      "step": 5200
+    },
+    {
+      "epoch": 2.7926960257787323,
+      "eval_accuracy": 0.9747583243823845,
+      "eval_loss": 0.15087123215198517,
+      "eval_runtime": 61.9759,
+      "eval_samples_per_second": 60.088,
+      "eval_steps_per_second": 7.519,
+      "step": 5200
+    },
+    {
+      "epoch": 2.9001074113856067,
+      "grad_norm": 0.013723284937441349,
+      "learning_rate": 2.752416756176155e-06,
+      "loss": 0.0055,
+      "step": 5400
+    },
+    {
+      "epoch": 2.9001074113856067,
+      "eval_accuracy": 0.9747583243823845,
+      "eval_loss": 0.16743087768554688,
+      "eval_runtime": 61.6792,
+      "eval_samples_per_second": 60.377,
+      "eval_steps_per_second": 7.555,
+      "step": 5400
+    },
+    {
+      "epoch": 3.007518796992481,
+      "grad_norm": 0.00023605262686032802,
+      "learning_rate": 2.483888292158969e-06,
+      "loss": 0.0094,
+      "step": 5600
+    },
+    {
+      "epoch": 3.007518796992481,
+      "eval_accuracy": 0.9747583243823845,
+      "eval_loss": 0.15689106285572052,
+      "eval_runtime": 62.4872,
+      "eval_samples_per_second": 59.596,
+      "eval_steps_per_second": 7.458,
+      "step": 5600
+    },
+    {
+      "epoch": 3.1149301825993554,
+      "grad_norm": 0.031025564298033714,
+      "learning_rate": 2.215359828141783e-06,
+      "loss": 0.0018,
+      "step": 5800
+    },
+    {
+      "epoch": 3.1149301825993554,
+      "eval_accuracy": 0.9752953813104189,
+      "eval_loss": 0.15795429050922394,
+      "eval_runtime": 61.1051,
+      "eval_samples_per_second": 60.944,
+      "eval_steps_per_second": 7.626,
+      "step": 5800
+    },
+    {
+      "epoch": 3.2223415682062297,
+      "grad_norm": 0.00042387741268612444,
+      "learning_rate": 1.9468313641245973e-06,
+      "loss": 0.0,
+      "step": 6000
+    },
+    {
+      "epoch": 3.2223415682062297,
+      "eval_accuracy": 0.9761009667024705,
+      "eval_loss": 0.16977229714393616,
+      "eval_runtime": 61.4203,
+      "eval_samples_per_second": 60.631,
+      "eval_steps_per_second": 7.587,
+      "step": 6000
+    },
+    {
+      "epoch": 3.329752953813104,
+      "grad_norm": 0.0005808643181808293,
+      "learning_rate": 1.6783029001074116e-06,
+      "loss": 0.0003,
+      "step": 6200
+    },
+    {
+      "epoch": 3.329752953813104,
+      "eval_accuracy": 0.9761009667024705,
+      "eval_loss": 0.1605655997991562,
+      "eval_runtime": 61.8621,
+      "eval_samples_per_second": 60.198,
+      "eval_steps_per_second": 7.533,
+      "step": 6200
+    },
+    {
+      "epoch": 3.4371643394199785,
+      "grad_norm": 5.516281453310512e-05,
+      "learning_rate": 1.4097744360902258e-06,
+      "loss": 0.0034,
+      "step": 6400
+    },
+    {
+      "epoch": 3.4371643394199785,
+      "eval_accuracy": 0.9736842105263158,
+      "eval_loss": 0.18701837956905365,
+      "eval_runtime": 61.6799,
+      "eval_samples_per_second": 60.376,
+      "eval_steps_per_second": 7.555,
+      "step": 6400
+    },
+    {
+      "epoch": 3.544575725026853,
+      "grad_norm": 0.005959503818303347,
+      "learning_rate": 1.1412459720730397e-06,
+      "loss": 0.0,
+      "step": 6600
+    },
+    {
+      "epoch": 3.544575725026853,
+      "eval_accuracy": 0.9755639097744361,
+      "eval_loss": 0.16974356770515442,
+      "eval_runtime": 61.5705,
+      "eval_samples_per_second": 60.483,
+      "eval_steps_per_second": 7.569,
+      "step": 6600
+    },
+    {
+      "epoch": 3.651987110633727,
+      "grad_norm": 0.0005512916832230985,
+      "learning_rate": 8.727175080558539e-07,
+      "loss": 0.0,
+      "step": 6800
+    },
+    {
+      "epoch": 3.651987110633727,
+      "eval_accuracy": 0.9750268528464017,
+      "eval_loss": 0.1672970950603485,
+      "eval_runtime": 61.3615,
+      "eval_samples_per_second": 60.69,
+      "eval_steps_per_second": 7.594,
+      "step": 6800
+    },
+    {
+      "epoch": 3.7593984962406015,
+      "grad_norm": 0.0007423324859701097,
+      "learning_rate": 6.041890440386682e-07,
+      "loss": 0.0053,
+      "step": 7000
+    },
+    {
+      "epoch": 3.7593984962406015,
+      "eval_accuracy": 0.9752953813104189,
+      "eval_loss": 0.16435429453849792,
+      "eval_runtime": 63.0316,
+      "eval_samples_per_second": 59.081,
+      "eval_steps_per_second": 7.393,
+      "step": 7000
+    },
+    {
+      "epoch": 3.866809881847476,
+      "grad_norm": 0.0015081085730344057,
+      "learning_rate": 3.356605800214823e-07,
+      "loss": 0.0,
+      "step": 7200
+    },
+    {
+      "epoch": 3.866809881847476,
+      "eval_accuracy": 0.9752953813104189,
+      "eval_loss": 0.16764488816261292,
+      "eval_runtime": 61.9177,
+      "eval_samples_per_second": 60.144,
+      "eval_steps_per_second": 7.526,
+      "step": 7200
+    },
+    {
+      "epoch": 3.9742212674543502,
+      "grad_norm": 0.00019354032701812685,
+      "learning_rate": 6.713211600429646e-08,
+      "loss": 0.0013,
+      "step": 7400
+    },
+    {
+      "epoch": 3.9742212674543502,
+      "eval_accuracy": 0.9761009667024705,
+      "eval_loss": 0.1640813946723938,
+      "eval_runtime": 61.6171,
+      "eval_samples_per_second": 60.438,
+      "eval_steps_per_second": 7.563,
+      "step": 7400
+    },
+    {
+      "epoch": 4.0,
+      "step": 7448,
+      "total_flos": 1.668149581703322e+19,
+      "train_loss": 0.03976387239464436,
+      "train_runtime": 7938.8524,
+      "train_samples_per_second": 7.504,
+      "train_steps_per_second": 0.938
     }
   ],
   "logging_steps": 200,
-  "max_steps": 2212,
+  "max_steps": 7448,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 2,
+  "num_train_epochs": 4,
   "save_steps": 200,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -211,8 +627,8 @@
       "attributes": {}
     }
   },
-  "total_flos": 9.690147678529511e+18,
-  "train_batch_size": 16,
+  "total_flos": 1.668149581703322e+19,
+  "train_batch_size": 8,
   "trial_name": null,
   "trial_params": null
 }