dhritic99/vit-base-brain-alzheimer-detection

README.md

@@ -17,8 +17,8 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1892
-- Accuracy: 0.9590
+- Loss: 0.2301
+- Accuracy: 0.9555
 
 ## Model description
 

all_results.json

@@ -1,13 +1,13 @@
 {
-    "epoch": 13.28125,
-    "eval_accuracy": 0.95078125,
-    "eval_loss": 0.24280217289924622,
-    "eval_runtime": 6.969,
-    "eval_samples_per_second": 183.67,
-    "eval_steps_per_second": 22.959,
-    "total_flos": 4.2156398005714944e+18,
-    "train_loss": 0.23371873112285838,
-    "train_runtime": 1043.0698,
-    "train_samples_per_second": 78.537,
-    "train_steps_per_second": 4.909
+    "epoch": 15.625,
+    "eval_accuracy": 0.95546875,
+    "eval_loss": 0.2301121950149536,
+    "eval_runtime": 7.6501,
+    "eval_samples_per_second": 167.318,
+    "eval_steps_per_second": 20.915,
+    "total_flos": 4.959576235966464e+18,
+    "train_loss": 0.10841592013835907,
+    "train_runtime": 976.7829,
+    "train_samples_per_second": 83.867,
+    "train_steps_per_second": 5.242
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 13.28125,
-    "eval_accuracy": 0.95078125,
-    "eval_loss": 0.24280217289924622,
-    "eval_runtime": 6.969,
-    "eval_samples_per_second": 183.67,
-    "eval_steps_per_second": 22.959
+    "epoch": 15.625,
+    "eval_accuracy": 0.95546875,
+    "eval_loss": 0.2301121950149536,
+    "eval_runtime": 7.6501,
+    "eval_samples_per_second": 167.318,
+    "eval_steps_per_second": 20.915
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 13.28125,
-    "total_flos": 4.2156398005714944e+18,
-    "train_loss": 0.23371873112285838,
-    "train_runtime": 1043.0698,
-    "train_samples_per_second": 78.537,
-    "train_steps_per_second": 4.909
+    "epoch": 15.625,
+    "total_flos": 4.959576235966464e+18,
+    "train_loss": 0.10841592013835907,
+    "train_runtime": 976.7829,
+    "train_samples_per_second": 83.867,
+    "train_steps_per_second": 5.242
 }

trainer_state.json

@@ -1,657 +1,660 @@
 {
-  "best_metric": 0.19315651059150696,
-  "best_model_checkpoint": "./vit-base-brain-alzheimer-detection/checkpoint-2800",
-  "epoch": 13.28125,
-  "eval_steps": 200,
-  "global_step": 3400,
+  "best_metric": 0.16233286261558533,
+  "best_model_checkpoint": "./vit-base-brain-alzheimer-detection/checkpoint-2500",
+  "epoch": 15.625,
+  "eval_steps": 500,
+  "global_step": 4000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.1953125,
-      "grad_norm": 3.182270050048828,
+      "grad_norm": 5.237834453582764,
       "learning_rate": 2.0000000000000003e-06,
-      "loss": 1.401,
+      "loss": 0.6443,
       "step": 50
     },
     {
       "epoch": 0.390625,
-      "grad_norm": 2.6782896518707275,
+      "grad_norm": 10.076339721679688,
       "learning_rate": 4.000000000000001e-06,
-      "loss": 1.1871,
+      "loss": 0.5744,
       "step": 100
     },
     {
       "epoch": 0.5859375,
-      "grad_norm": 2.7876715660095215,
+      "grad_norm": 10.711421012878418,
       "learning_rate": 6e-06,
-      "loss": 1.035,
+      "loss": 0.5509,
       "step": 150
     },
     {
       "epoch": 0.78125,
-      "grad_norm": 3.130065441131592,
+      "grad_norm": 7.218406677246094,
       "learning_rate": 8.000000000000001e-06,
-      "loss": 0.9772,
-      "step": 200
-    },
-    {
-      "epoch": 0.78125,
-      "eval_accuracy": 0.580078125,
-      "eval_loss": 0.9400424957275391,
-      "eval_runtime": 5.935,
-      "eval_samples_per_second": 172.535,
-      "eval_steps_per_second": 21.567,
+      "loss": 0.5198,
       "step": 200
     },
     {
       "epoch": 0.9765625,
-      "grad_norm": 2.5494868755340576,
+      "grad_norm": 3.6985344886779785,
       "learning_rate": 1e-05,
-      "loss": 0.8886,
+      "loss": 0.4411,
       "step": 250
     },
     {
       "epoch": 1.171875,
-      "grad_norm": 1.62320876121521,
+      "grad_norm": 29.415699005126953,
       "learning_rate": 1.2e-05,
-      "loss": 0.8722,
+      "loss": 0.4885,
       "step": 300
     },
     {
       "epoch": 1.3671875,
-      "grad_norm": 4.327115535736084,
+      "grad_norm": 24.613481521606445,
       "learning_rate": 1.4e-05,
-      "loss": 0.8154,
+      "loss": 0.4818,
       "step": 350
     },
     {
       "epoch": 1.5625,
-      "grad_norm": 7.814743995666504,
+      "grad_norm": 8.766143798828125,
       "learning_rate": 1.6000000000000003e-05,
-      "loss": 0.7451,
-      "step": 400
-    },
-    {
-      "epoch": 1.5625,
-      "eval_accuracy": 0.6552734375,
-      "eval_loss": 0.7946789264678955,
-      "eval_runtime": 7.068,
-      "eval_samples_per_second": 144.879,
-      "eval_steps_per_second": 18.11,
+      "loss": 0.4116,
       "step": 400
     },
     {
       "epoch": 1.7578125,
-      "grad_norm": 15.593164443969727,
+      "grad_norm": 2.4442567825317383,
       "learning_rate": 1.8e-05,
-      "loss": 0.7383,
+      "loss": 0.4328,
       "step": 450
     },
     {
       "epoch": 1.953125,
-      "grad_norm": 5.583272933959961,
+      "grad_norm": 3.787712335586548,
       "learning_rate": 2e-05,
-      "loss": 0.733,
+      "loss": 0.4285,
+      "step": 500
+    },
+    {
+      "epoch": 1.953125,
+      "eval_accuracy": 0.8310546875,
+      "eval_loss": 0.4633047878742218,
+      "eval_runtime": 5.4233,
+      "eval_samples_per_second": 188.814,
+      "eval_steps_per_second": 23.602,
       "step": 500
     },
     {
       "epoch": 2.1484375,
-      "grad_norm": 10.433585166931152,
+      "grad_norm": 2.96122407913208,
       "learning_rate": 1.9783549783549785e-05,
-      "loss": 0.5905,
+      "loss": 0.3494,
       "step": 550
     },
     {
       "epoch": 2.34375,
-      "grad_norm": 4.751012802124023,
+      "grad_norm": 4.47251558303833,
       "learning_rate": 1.9567099567099568e-05,
-      "loss": 0.5701,
-      "step": 600
-    },
-    {
-      "epoch": 2.34375,
-      "eval_accuracy": 0.7236328125,
-      "eval_loss": 0.7641921639442444,
-      "eval_runtime": 5.5404,
-      "eval_samples_per_second": 184.824,
-      "eval_steps_per_second": 23.103,
+      "loss": 0.3011,
       "step": 600
     },
     {
       "epoch": 2.5390625,
-      "grad_norm": 17.00213050842285,
+      "grad_norm": 6.335562229156494,
       "learning_rate": 1.9350649350649354e-05,
-      "loss": 0.6235,
+      "loss": 0.3097,
       "step": 650
     },
     {
       "epoch": 2.734375,
-      "grad_norm": 10.933674812316895,
+      "grad_norm": 3.6047630310058594,
       "learning_rate": 1.9134199134199138e-05,
-      "loss": 0.5385,
+      "loss": 0.2601,
       "step": 700
     },
     {
       "epoch": 2.9296875,
-      "grad_norm": 9.459794044494629,
+      "grad_norm": 13.996108055114746,
       "learning_rate": 1.891774891774892e-05,
-      "loss": 0.4557,
+      "loss": 0.3092,
       "step": 750
     },
     {
       "epoch": 3.125,
-      "grad_norm": 2.7128560543060303,
+      "grad_norm": 5.770679473876953,
       "learning_rate": 1.8701298701298704e-05,
-      "loss": 0.3704,
-      "step": 800
-    },
-    {
-      "epoch": 3.125,
-      "eval_accuracy": 0.7744140625,
-      "eval_loss": 0.5531814098358154,
-      "eval_runtime": 6.1892,
-      "eval_samples_per_second": 165.45,
-      "eval_steps_per_second": 20.681,
+      "loss": 0.2201,
       "step": 800
     },
     {
       "epoch": 3.3203125,
-      "grad_norm": 6.759947776794434,
+      "grad_norm": 20.05348014831543,
       "learning_rate": 1.8484848484848487e-05,
-      "loss": 0.3876,
+      "loss": 0.2323,
       "step": 850
     },
     {
       "epoch": 3.515625,
-      "grad_norm": 4.957170009613037,
+      "grad_norm": 1.26289963722229,
       "learning_rate": 1.826839826839827e-05,
-      "loss": 0.3014,
+      "loss": 0.1483,
       "step": 900
     },
     {
       "epoch": 3.7109375,
-      "grad_norm": 4.334908485412598,
+      "grad_norm": 2.166567087173462,
       "learning_rate": 1.8051948051948053e-05,
-      "loss": 0.2523,
+      "loss": 0.2116,
       "step": 950
     },
     {
       "epoch": 3.90625,
-      "grad_norm": 12.843859672546387,
+      "grad_norm": 25.671602249145508,
       "learning_rate": 1.7835497835497836e-05,
-      "loss": 0.2906,
+      "loss": 0.171,
       "step": 1000
     },
     {
       "epoch": 3.90625,
-      "eval_accuracy": 0.85546875,
-      "eval_loss": 0.4422951340675354,
-      "eval_runtime": 5.5363,
-      "eval_samples_per_second": 184.962,
-      "eval_steps_per_second": 23.12,
+      "eval_accuracy": 0.8994140625,
+      "eval_loss": 0.32369568943977356,
+      "eval_runtime": 5.6226,
+      "eval_samples_per_second": 182.123,
+      "eval_steps_per_second": 22.765,
       "step": 1000
     },
     {
       "epoch": 4.1015625,
-      "grad_norm": 3.9840188026428223,
+      "grad_norm": 1.2532105445861816,
       "learning_rate": 1.761904761904762e-05,
-      "loss": 0.2216,
+      "loss": 0.159,
       "step": 1050
     },
     {
       "epoch": 4.296875,
-      "grad_norm": 11.730119705200195,
+      "grad_norm": 0.9766639471054077,
       "learning_rate": 1.7402597402597403e-05,
-      "loss": 0.1722,
+      "loss": 0.1412,
       "step": 1100
     },
     {
       "epoch": 4.4921875,
-      "grad_norm": 2.089110851287842,
+      "grad_norm": 2.724047899246216,
       "learning_rate": 1.718614718614719e-05,
-      "loss": 0.1427,
+      "loss": 0.082,
       "step": 1150
     },
     {
       "epoch": 4.6875,
-      "grad_norm": 0.5950905084609985,
+      "grad_norm": 1.2899634838104248,
       "learning_rate": 1.6969696969696972e-05,
-      "loss": 0.1636,
-      "step": 1200
-    },
-    {
-      "epoch": 4.6875,
-      "eval_accuracy": 0.900390625,
-      "eval_loss": 0.32256966829299927,
-      "eval_runtime": 6.2703,
-      "eval_samples_per_second": 163.31,
-      "eval_steps_per_second": 20.414,
+      "loss": 0.0837,
       "step": 1200
     },
     {
       "epoch": 4.8828125,
-      "grad_norm": 4.693203926086426,
+      "grad_norm": 2.9213526248931885,
       "learning_rate": 1.6753246753246756e-05,
-      "loss": 0.1622,
+      "loss": 0.0554,
       "step": 1250
     },
     {
       "epoch": 5.078125,
-      "grad_norm": 0.24195541441440582,
+      "grad_norm": 0.10849986970424652,
       "learning_rate": 1.653679653679654e-05,
-      "loss": 0.099,
+      "loss": 0.0681,
       "step": 1300
     },
     {
       "epoch": 5.2734375,
-      "grad_norm": 0.16812123358249664,
+      "grad_norm": 0.07504723966121674,
       "learning_rate": 1.6320346320346322e-05,
-      "loss": 0.0979,
+      "loss": 0.0415,
       "step": 1350
     },
     {
       "epoch": 5.46875,
-      "grad_norm": 19.697050094604492,
+      "grad_norm": 0.1930815428495407,
       "learning_rate": 1.6103896103896105e-05,
-      "loss": 0.0837,
-      "step": 1400
-    },
-    {
-      "epoch": 5.46875,
-      "eval_accuracy": 0.90234375,
-      "eval_loss": 0.3483480215072632,
-      "eval_runtime": 5.5142,
-      "eval_samples_per_second": 185.702,
-      "eval_steps_per_second": 23.213,
+      "loss": 0.0333,
       "step": 1400
     },
     {
       "epoch": 5.6640625,
-      "grad_norm": 0.35744136571884155,
+      "grad_norm": 0.41994306445121765,
       "learning_rate": 1.5887445887445888e-05,
-      "loss": 0.118,
+      "loss": 0.0402,
       "step": 1450
     },
     {
       "epoch": 5.859375,
-      "grad_norm": 0.1951870173215866,
+      "grad_norm": 0.08935032039880753,
       "learning_rate": 1.567099567099567e-05,
-      "loss": 0.0826,
+      "loss": 0.0622,
+      "step": 1500
+    },
+    {
+      "epoch": 5.859375,
+      "eval_accuracy": 0.94140625,
+      "eval_loss": 0.20316675305366516,
+      "eval_runtime": 5.9862,
+      "eval_samples_per_second": 171.06,
+      "eval_steps_per_second": 21.382,
       "step": 1500
     },
     {
       "epoch": 6.0546875,
-      "grad_norm": 27.74083137512207,
+      "grad_norm": 0.08510852605104446,
       "learning_rate": 1.5454545454545454e-05,
-      "loss": 0.0697,
+      "loss": 0.0407,
       "step": 1550
     },
     {
       "epoch": 6.25,
-      "grad_norm": 1.9396530389785767,
+      "grad_norm": 12.273158073425293,
       "learning_rate": 1.523809523809524e-05,
-      "loss": 0.0368,
-      "step": 1600
-    },
-    {
-      "epoch": 6.25,
-      "eval_accuracy": 0.939453125,
-      "eval_loss": 0.24231916666030884,
-      "eval_runtime": 5.9487,
-      "eval_samples_per_second": 172.139,
-      "eval_steps_per_second": 21.517,
+      "loss": 0.0207,
       "step": 1600
     },
     {
       "epoch": 6.4453125,
-      "grad_norm": 0.1338203400373459,
+      "grad_norm": 0.099972203373909,
       "learning_rate": 1.5021645021645024e-05,
-      "loss": 0.0482,
+      "loss": 0.0756,
       "step": 1650
     },
     {
       "epoch": 6.640625,
-      "grad_norm": 0.16084209084510803,
+      "grad_norm": 0.20563142001628876,
       "learning_rate": 1.4805194805194807e-05,
-      "loss": 0.0382,
+      "loss": 0.0277,
       "step": 1700
     },
     {
       "epoch": 6.8359375,
-      "grad_norm": 13.326003074645996,
+      "grad_norm": 0.0686117634177208,
       "learning_rate": 1.458874458874459e-05,
-      "loss": 0.1169,
+      "loss": 0.0167,
       "step": 1750
     },
     {
       "epoch": 7.03125,
-      "grad_norm": 0.1118919774889946,
+      "grad_norm": 0.036712419241666794,
       "learning_rate": 1.4372294372294374e-05,
-      "loss": 0.063,
-      "step": 1800
-    },
-    {
-      "epoch": 7.03125,
-      "eval_accuracy": 0.927734375,
-      "eval_loss": 0.30914944410324097,
-      "eval_runtime": 5.5136,
-      "eval_samples_per_second": 185.722,
-      "eval_steps_per_second": 23.215,
+      "loss": 0.0174,
       "step": 1800
     },
     {
       "epoch": 7.2265625,
-      "grad_norm": 0.11637281626462936,
+      "grad_norm": 0.04611894115805626,
       "learning_rate": 1.4155844155844157e-05,
-      "loss": 0.0385,
+      "loss": 0.0161,
       "step": 1850
     },
     {
       "epoch": 7.421875,
-      "grad_norm": 0.09042005985975266,
+      "grad_norm": 0.0400649756193161,
       "learning_rate": 1.3939393939393942e-05,
-      "loss": 0.0277,
+      "loss": 0.0087,
       "step": 1900
     },
     {
       "epoch": 7.6171875,
-      "grad_norm": 0.07495836168527603,
+      "grad_norm": 0.036735936999320984,
       "learning_rate": 1.3722943722943725e-05,
-      "loss": 0.0214,
+      "loss": 0.0083,
       "step": 1950
     },
     {
       "epoch": 7.8125,
-      "grad_norm": 0.06756331771612167,
+      "grad_norm": 0.031237976625561714,
       "learning_rate": 1.3506493506493508e-05,
-      "loss": 0.047,
+      "loss": 0.0162,
       "step": 2000
     },
     {
       "epoch": 7.8125,
-      "eval_accuracy": 0.90234375,
-      "eval_loss": 0.3906704783439636,
-      "eval_runtime": 6.0843,
-      "eval_samples_per_second": 168.303,
-      "eval_steps_per_second": 21.038,
+      "eval_accuracy": 0.951171875,
+      "eval_loss": 0.241306334733963,
+      "eval_runtime": 5.4826,
+      "eval_samples_per_second": 186.771,
+      "eval_steps_per_second": 23.346,
       "step": 2000
     },
     {
       "epoch": 8.0078125,
-      "grad_norm": 0.3716655671596527,
-      "learning_rate": 1.3290043290043291e-05,
-      "loss": 0.0262,
+      "grad_norm": 0.027786873281002045,
+      "learning_rate": 1.3294372294372295e-05,
+      "loss": 0.0162,
       "step": 2050
     },
     {
       "epoch": 8.203125,
-      "grad_norm": 0.40872329473495483,
-      "learning_rate": 1.3073593073593074e-05,
-      "loss": 0.0139,
+      "grad_norm": 0.06131181865930557,
+      "learning_rate": 1.3077922077922078e-05,
+      "loss": 0.0085,
       "step": 2100
     },
     {
       "epoch": 8.3984375,
-      "grad_norm": 0.05600840970873833,
+      "grad_norm": 0.044134218245744705,
       "learning_rate": 1.2861471861471862e-05,
-      "loss": 0.0161,
+      "loss": 0.0152,
       "step": 2150
     },
     {
       "epoch": 8.59375,
-      "grad_norm": 0.055402472615242004,
+      "grad_norm": 10.187455177307129,
       "learning_rate": 1.2645021645021646e-05,
-      "loss": 0.0127,
-      "step": 2200
-    },
-    {
-      "epoch": 8.59375,
-      "eval_accuracy": 0.9560546875,
-      "eval_loss": 0.20022055506706238,
-      "eval_runtime": 5.6277,
-      "eval_samples_per_second": 181.956,
-      "eval_steps_per_second": 22.744,
+      "loss": 0.0167,
       "step": 2200
     },
     {
       "epoch": 8.7890625,
-      "grad_norm": 0.054736897349357605,
+      "grad_norm": 0.024555057287216187,
       "learning_rate": 1.242857142857143e-05,
-      "loss": 0.0135,
+      "loss": 0.0092,
       "step": 2250
     },
     {
       "epoch": 8.984375,
-      "grad_norm": 0.04909529909491539,
+      "grad_norm": 1.3849960565567017,
       "learning_rate": 1.2212121212121213e-05,
-      "loss": 0.0209,
+      "loss": 0.0068,
       "step": 2300
     },
     {
       "epoch": 9.1796875,
-      "grad_norm": 0.04906412959098816,
+      "grad_norm": 0.021510746330022812,
       "learning_rate": 1.1995670995670996e-05,
-      "loss": 0.0107,
+      "loss": 0.0051,
       "step": 2350
     },
     {
       "epoch": 9.375,
-      "grad_norm": 0.04470803588628769,
+      "grad_norm": 0.01960117556154728,
       "learning_rate": 1.1779220779220779e-05,
-      "loss": 0.0102,
-      "step": 2400
-    },
-    {
-      "epoch": 9.375,
-      "eval_accuracy": 0.9306640625,
-      "eval_loss": 0.30005937814712524,
-      "eval_runtime": 6.136,
-      "eval_samples_per_second": 166.885,
-      "eval_steps_per_second": 20.861,
+      "loss": 0.0045,
       "step": 2400
     },
     {
       "epoch": 9.5703125,
-      "grad_norm": 0.04809673875570297,
+      "grad_norm": 0.021382611244916916,
       "learning_rate": 1.1562770562770564e-05,
-      "loss": 0.01,
+      "loss": 0.0046,
       "step": 2450
     },
     {
       "epoch": 9.765625,
-      "grad_norm": 0.053658001124858856,
+      "grad_norm": 0.031110389158129692,
       "learning_rate": 1.1346320346320347e-05,
-      "loss": 0.0096,
+      "loss": 0.0044,
+      "step": 2500
+    },
+    {
+      "epoch": 9.765625,
+      "eval_accuracy": 0.966796875,
+      "eval_loss": 0.16233286261558533,
+      "eval_runtime": 5.8021,
+      "eval_samples_per_second": 176.487,
+      "eval_steps_per_second": 22.061,
       "step": 2500
     },
     {
       "epoch": 9.9609375,
-      "grad_norm": 0.05094292759895325,
+      "grad_norm": 0.01663753017783165,
       "learning_rate": 1.112987012987013e-05,
-      "loss": 0.0085,
+      "loss": 0.0039,
       "step": 2550
     },
     {
       "epoch": 10.15625,
-      "grad_norm": 0.03785080462694168,
+      "grad_norm": 0.016748279333114624,
       "learning_rate": 1.0913419913419913e-05,
-      "loss": 0.0086,
-      "step": 2600
-    },
-    {
-      "epoch": 10.15625,
-      "eval_accuracy": 0.951171875,
-      "eval_loss": 0.19980870187282562,
-      "eval_runtime": 5.5065,
-      "eval_samples_per_second": 185.964,
-      "eval_steps_per_second": 23.245,
+      "loss": 0.004,
       "step": 2600
     },
     {
       "epoch": 10.3515625,
-      "grad_norm": 0.048504624515771866,
+      "grad_norm": 0.019847417250275612,
       "learning_rate": 1.0696969696969696e-05,
-      "loss": 0.0078,
+      "loss": 0.0036,
       "step": 2650
     },
     {
       "epoch": 10.546875,
-      "grad_norm": 0.040160298347473145,
+      "grad_norm": 0.01919238083064556,
       "learning_rate": 1.0480519480519481e-05,
-      "loss": 0.0136,
+      "loss": 0.0037,
       "step": 2700
     },
     {
       "epoch": 10.7421875,
-      "grad_norm": 0.03454400226473808,
+      "grad_norm": 0.01820230484008789,
       "learning_rate": 1.0264069264069264e-05,
-      "loss": 0.0077,
+      "loss": 0.0036,
       "step": 2750
     },
     {
       "epoch": 10.9375,
-      "grad_norm": 0.03354168310761452,
+      "grad_norm": 0.01602366380393505,
       "learning_rate": 1.0047619047619048e-05,
-      "loss": 0.0073,
-      "step": 2800
-    },
-    {
-      "epoch": 10.9375,
-      "eval_accuracy": 0.958984375,
-      "eval_loss": 0.19315651059150696,
-      "eval_runtime": 6.0024,
-      "eval_samples_per_second": 170.598,
-      "eval_steps_per_second": 21.325,
+      "loss": 0.0034,
       "step": 2800
     },
     {
       "epoch": 11.1328125,
-      "grad_norm": 0.033880043774843216,
+      "grad_norm": 0.01528051309287548,
       "learning_rate": 9.831168831168832e-06,
-      "loss": 0.0071,
+      "loss": 0.0033,
       "step": 2850
     },
     {
       "epoch": 11.328125,
-      "grad_norm": 0.032037124037742615,
+      "grad_norm": 0.015123172663152218,
       "learning_rate": 9.614718614718616e-06,
-      "loss": 0.0066,
+      "loss": 0.003,
       "step": 2900
     },
     {
       "epoch": 11.5234375,
-      "grad_norm": 0.0329328328371048,
+      "grad_norm": 0.017041392624378204,
       "learning_rate": 9.398268398268399e-06,
-      "loss": 0.0066,
+      "loss": 0.0031,
       "step": 2950
     },
     {
       "epoch": 11.71875,
-      "grad_norm": 0.02779380977153778,
+      "grad_norm": 0.012525764293968678,
       "learning_rate": 9.181818181818184e-06,
-      "loss": 0.0064,
+      "loss": 0.003,
       "step": 3000
     },
     {
       "epoch": 11.71875,
-      "eval_accuracy": 0.9560546875,
-      "eval_loss": 0.19875724613666534,
-      "eval_runtime": 5.8842,
-      "eval_samples_per_second": 174.025,
-      "eval_steps_per_second": 21.753,
+      "eval_accuracy": 0.966796875,
+      "eval_loss": 0.16406887769699097,
+      "eval_runtime": 6.0958,
+      "eval_samples_per_second": 167.985,
+      "eval_steps_per_second": 20.998,
       "step": 3000
     },
     {
       "epoch": 11.9140625,
-      "grad_norm": 0.028905285522341728,
+      "grad_norm": 0.014454510062932968,
       "learning_rate": 8.965367965367967e-06,
-      "loss": 0.0063,
+      "loss": 0.003,
       "step": 3050
     },
     {
       "epoch": 12.109375,
-      "grad_norm": 0.027350405231118202,
+      "grad_norm": 0.012859505601227283,
       "learning_rate": 8.74891774891775e-06,
-      "loss": 0.0058,
+      "loss": 0.0027,
       "step": 3100
     },
     {
       "epoch": 12.3046875,
-      "grad_norm": 0.026442719623446465,
+      "grad_norm": 0.012520495802164078,
       "learning_rate": 8.532467532467533e-06,
-      "loss": 0.0059,
+      "loss": 0.0028,
       "step": 3150
     },
     {
       "epoch": 12.5,
-      "grad_norm": 0.028089182451367378,
+      "grad_norm": 0.014481940306723118,
       "learning_rate": 8.316017316017316e-06,
-      "loss": 0.0056,
-      "step": 3200
-    },
-    {
-      "epoch": 12.5,
-      "eval_accuracy": 0.9580078125,
-      "eval_loss": 0.19929003715515137,
-      "eval_runtime": 6.1143,
-      "eval_samples_per_second": 167.475,
-      "eval_steps_per_second": 20.934,
+      "loss": 0.0026,
       "step": 3200
     },
     {
       "epoch": 12.6953125,
-      "grad_norm": 0.02532665804028511,
+      "grad_norm": 0.01108886580914259,
       "learning_rate": 8.099567099567101e-06,
-      "loss": 0.0053,
+      "loss": 0.0025,
       "step": 3250
     },
     {
       "epoch": 12.890625,
-      "grad_norm": 0.024192778393626213,
+      "grad_norm": 0.011388474144041538,
       "learning_rate": 7.883116883116884e-06,
-      "loss": 0.0055,
+      "loss": 0.0026,
       "step": 3300
     },
     {
       "epoch": 13.0859375,
-      "grad_norm": 0.028951533138751984,
+      "grad_norm": 0.01610645279288292,
       "learning_rate": 7.666666666666667e-06,
-      "loss": 0.0053,
+      "loss": 0.0025,
       "step": 3350
     },
     {
       "epoch": 13.28125,
-      "grad_norm": 0.024818725883960724,
+      "grad_norm": 0.013085401616990566,
       "learning_rate": 7.450216450216451e-06,
-      "loss": 0.0049,
+      "loss": 0.0023,
       "step": 3400
     },
     {
-      "epoch": 13.28125,
-      "eval_accuracy": 0.958984375,
-      "eval_loss": 0.2046782374382019,
-      "eval_runtime": 5.5223,
-      "eval_samples_per_second": 185.43,
-      "eval_steps_per_second": 23.179,
-      "step": 3400
+      "epoch": 13.4765625,
+      "grad_norm": 0.012581477873027325,
+      "learning_rate": 7.233766233766234e-06,
+      "loss": 0.0023,
+      "step": 3450
     },
     {
-      "epoch": 13.28125,
-      "step": 3400,
-      "total_flos": 4.2156398005714944e+18,
-      "train_loss": 0.23371873112285838,
-      "train_runtime": 1043.0698,
-      "train_samples_per_second": 78.537,
-      "train_steps_per_second": 4.909
+      "epoch": 13.671875,
+      "grad_norm": 0.01144211646169424,
+      "learning_rate": 7.017316017316018e-06,
+      "loss": 0.0025,
+      "step": 3500
+    },
+    {
+      "epoch": 13.671875,
+      "eval_accuracy": 0.9619140625,
+      "eval_loss": 0.17963293194770813,
+      "eval_runtime": 6.1937,
+      "eval_samples_per_second": 165.331,
+      "eval_steps_per_second": 20.666,
+      "step": 3500
+    },
+    {
+      "epoch": 13.8671875,
+      "grad_norm": 0.01094707939773798,
+      "learning_rate": 6.8008658008658015e-06,
+      "loss": 0.0022,
+      "step": 3550
+    },
+    {
+      "epoch": 14.0625,
+      "grad_norm": 0.010639617219567299,
+      "learning_rate": 6.5844155844155855e-06,
+      "loss": 0.0022,
+      "step": 3600
+    },
+    {
+      "epoch": 14.2578125,
+      "grad_norm": 0.02006010338664055,
+      "learning_rate": 6.367965367965369e-06,
+      "loss": 0.0021,
+      "step": 3650
+    },
+    {
+      "epoch": 14.453125,
+      "grad_norm": 0.008836320601403713,
+      "learning_rate": 6.151515151515152e-06,
+      "loss": 0.0021,
+      "step": 3700
+    },
+    {
+      "epoch": 14.6484375,
+      "grad_norm": 0.009966148994863033,
+      "learning_rate": 5.935064935064936e-06,
+      "loss": 0.0021,
+      "step": 3750
+    },
+    {
+      "epoch": 14.84375,
+      "grad_norm": 0.00935059692710638,
+      "learning_rate": 5.718614718614719e-06,
+      "loss": 0.0021,
+      "step": 3800
+    },
+    {
+      "epoch": 15.0390625,
+      "grad_norm": 0.009346270933747292,
+      "learning_rate": 5.502164502164503e-06,
+      "loss": 0.002,
+      "step": 3850
+    },
+    {
+      "epoch": 15.234375,
+      "grad_norm": 0.00850555207580328,
+      "learning_rate": 5.285714285714286e-06,
+      "loss": 0.0019,
+      "step": 3900
+    },
+    {
+      "epoch": 15.4296875,
+      "grad_norm": 0.009360904805362225,
+      "learning_rate": 5.069264069264069e-06,
+      "loss": 0.002,
+      "step": 3950
+    },
+    {
+      "epoch": 15.625,
+      "grad_norm": 0.00921907089650631,
+      "learning_rate": 4.852813852813853e-06,
+      "loss": 0.0019,
+      "step": 4000
+    },
+    {
+      "epoch": 15.625,
+      "eval_accuracy": 0.958984375,
+      "eval_loss": 0.18917381763458252,
+      "eval_runtime": 5.6954,
+      "eval_samples_per_second": 179.796,
+      "eval_steps_per_second": 22.474,
+      "step": 4000
+    },
+    {
+      "epoch": 15.625,
+      "step": 4000,
+      "total_flos": 4.959576235966464e+18,
+      "train_loss": 0.10841592013835907,
+      "train_runtime": 976.7829,
+      "train_samples_per_second": 83.867,
+      "train_steps_per_second": 5.242
     }
   ],
   "logging_steps": 50,
   "max_steps": 5120,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 20,
-  "save_steps": 200,
+  "save_steps": 500,
   "stateful_callbacks": {
     "EarlyStoppingCallback": {
       "args": {
@@ -673,7 +676,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 4.2156398005714944e+18,
+  "total_flos": 4.959576235966464e+18,
   "train_batch_size": 16,
   "trial_name": null,
   "trial_params": null