End of training

all_results.json

@@ -1,13 +1,13 @@
 {
     "epoch": 1.0,
-    "eval_accuracy": 0.6422849181469871,
-    "eval_loss": 0.9593217372894287,
-    "eval_runtime": 25.0315,
-    "eval_samples_per_second": 114.696,
-    "eval_steps_per_second": 3.595,
-    "total_flos": 2.0014719749914337e+18,
-    "train_loss": 1.316133418885788,
-    "train_runtime": 615.4563,
-    "train_samples_per_second": 41.982,
-    "train_steps_per_second": 0.328
+    "eval_accuracy": 0.6415436054611312,
+    "eval_loss": 0.9517616629600525,
+    "eval_runtime": 62.4065,
+    "eval_samples_per_second": 115.02,
+    "eval_steps_per_second": 3.605,
+    "total_flos": 2.2210002502885048e+18,
+    "train_loss": 1.2996089096580232,
+    "train_runtime": 698.1963,
+    "train_samples_per_second": 41.119,
+    "train_steps_per_second": 0.321
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "eval_accuracy": 0.6422849181469871,
-    "eval_loss": 0.9593217372894287,
-    "eval_runtime": 25.0315,
-    "eval_samples_per_second": 114.696,
-    "eval_steps_per_second": 3.595
+    "eval_accuracy": 0.6415436054611312,
+    "eval_loss": 0.9517616629600525,
+    "eval_runtime": 62.4065,
+    "eval_samples_per_second": 115.02,
+    "eval_steps_per_second": 3.605
 }

runs/Jan08_22-21-08_2cd0117f6e75/events.out.tfevents.1673217328.2cd0117f6e75.24.5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c2a5bb364531ab218bb8077b343a59f1beee74d13c0fe596f912d2d1dd06778
+size 363

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 1.0,
-    "total_flos": 2.0014719749914337e+18,
-    "train_loss": 1.316133418885788,
-    "train_runtime": 615.4563,
-    "train_samples_per_second": 41.982,
-    "train_steps_per_second": 0.328
+    "total_flos": 2.2210002502885048e+18,
+    "train_loss": 1.2996089096580232,
+    "train_runtime": 698.1963,
+    "train_samples_per_second": 41.119,
+    "train_steps_per_second": 0.321
 }

trainer_state.json

@@ -1,154 +1,166 @@
 {
-  "best_metric": 0.6422849181469871,
-  "best_model_checkpoint": "beit-base-patch16-224-pt22k-ft22k-finetuned-FER2013-9e-05/checkpoint-202",
-  "epoch": 1.0,
-  "global_step": 202,
+  "best_metric": 0.6415436054611312,
+  "best_model_checkpoint": "beit-base-patch16-224-pt22k-ft22k-finetuned-FER2013-9e-05/checkpoint-224",
+  "epoch": 0.9977728285077951,
+  "global_step": 224,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.05,
-      "learning_rate": 4.2857142857142856e-05,
-      "loss": 1.9259,
+      "epoch": 0.04,
+      "learning_rate": 3.91304347826087e-05,
+      "loss": 1.9178,
       "step": 10
     },
     {
-      "epoch": 0.1,
-      "learning_rate": 8.571428571428571e-05,
-      "loss": 1.6569,
+      "epoch": 0.09,
+      "learning_rate": 7.82608695652174e-05,
+      "loss": 1.6833,
       "step": 20
     },
     {
-      "epoch": 0.15,
-      "learning_rate": 8.552486187845305e-05,
-      "loss": 1.4853,
+      "epoch": 0.13,
+      "learning_rate": 8.686567164179105e-05,
+      "loss": 1.5051,
       "step": 30
     },
     {
-      "epoch": 0.2,
-      "learning_rate": 8.055248618784531e-05,
-      "loss": 1.512,
+      "epoch": 0.18,
+      "learning_rate": 8.238805970149254e-05,
+      "loss": 1.4696,
       "step": 40
     },
     {
-      "epoch": 0.25,
-      "learning_rate": 7.558011049723757e-05,
-      "loss": 1.395,
+      "epoch": 0.22,
+      "learning_rate": 7.791044776119403e-05,
+      "loss": 1.3784,
       "step": 50
     },
     {
-      "epoch": 0.3,
-      "learning_rate": 7.060773480662983e-05,
-      "loss": 1.3225,
+      "epoch": 0.27,
+      "learning_rate": 7.343283582089552e-05,
+      "loss": 1.3547,
       "step": 60
     },
     {
-      "epoch": 0.35,
-      "learning_rate": 6.563535911602211e-05,
-      "loss": 1.3529,
+      "epoch": 0.31,
+      "learning_rate": 6.895522388059701e-05,
+      "loss": 1.3355,
       "step": 70
     },
     {
-      "epoch": 0.4,
-      "learning_rate": 6.0662983425414375e-05,
-      "loss": 1.3234,
+      "epoch": 0.36,
+      "learning_rate": 6.447761194029851e-05,
+      "loss": 1.2555,
       "step": 80
     },
     {
-      "epoch": 0.45,
-      "learning_rate": 5.569060773480663e-05,
-      "loss": 1.2544,
+      "epoch": 0.4,
+      "learning_rate": 6e-05,
+      "loss": 1.2671,
       "step": 90
     },
     {
-      "epoch": 0.5,
-      "learning_rate": 5.07182320441989e-05,
-      "loss": 1.2481,
+      "epoch": 0.45,
+      "learning_rate": 5.5522388059701494e-05,
+      "loss": 1.2823,
       "step": 100
     },
     {
-      "epoch": 0.54,
-      "learning_rate": 4.5745856353591164e-05,
-      "loss": 1.2172,
+      "epoch": 0.49,
+      "learning_rate": 5.1044776119402986e-05,
+      "loss": 1.2462,
       "step": 110
     },
     {
-      "epoch": 0.59,
-      "learning_rate": 4.077348066298343e-05,
-      "loss": 1.249,
+      "epoch": 0.53,
+      "learning_rate": 4.656716417910448e-05,
+      "loss": 1.207,
       "step": 120
     },
     {
-      "epoch": 0.64,
-      "learning_rate": 3.580110497237569e-05,
-      "loss": 1.2136,
+      "epoch": 0.58,
+      "learning_rate": 4.208955223880597e-05,
+      "loss": 1.1757,
       "step": 130
     },
     {
-      "epoch": 0.69,
-      "learning_rate": 3.082872928176796e-05,
-      "loss": 1.1581,
+      "epoch": 0.62,
+      "learning_rate": 3.761194029850746e-05,
+      "loss": 1.2294,
       "step": 140
     },
     {
-      "epoch": 0.74,
-      "learning_rate": 2.585635359116022e-05,
-      "loss": 1.1803,
+      "epoch": 0.67,
+      "learning_rate": 3.3134328358208955e-05,
+      "loss": 1.1943,
       "step": 150
     },
     {
-      "epoch": 0.79,
-      "learning_rate": 2.0883977900552486e-05,
-      "loss": 1.2273,
+      "epoch": 0.71,
+      "learning_rate": 2.865671641791045e-05,
+      "loss": 1.233,
       "step": 160
     },
     {
-      "epoch": 0.84,
-      "learning_rate": 1.591160220994475e-05,
-      "loss": 1.1976,
+      "epoch": 0.76,
+      "learning_rate": 2.4179104477611943e-05,
+      "loss": 1.1816,
       "step": 170
     },
     {
-      "epoch": 0.89,
-      "learning_rate": 1.0939226519337016e-05,
-      "loss": 1.1569,
+      "epoch": 0.8,
+      "learning_rate": 1.9701492537313435e-05,
+      "loss": 1.1394,
       "step": 180
     },
     {
-      "epoch": 0.94,
-      "learning_rate": 5.966850828729282e-06,
-      "loss": 1.178,
+      "epoch": 0.85,
+      "learning_rate": 1.5223880597014927e-05,
+      "loss": 1.1697,
       "step": 190
     },
     {
-      "epoch": 0.99,
-      "learning_rate": 9.944751381215469e-07,
-      "loss": 1.1404,
+      "epoch": 0.89,
+      "learning_rate": 1.0746268656716418e-05,
+      "loss": 1.1656,
       "step": 200
     },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.268656716417911e-06,
+      "loss": 1.1379,
+      "step": 210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7910447761194032e-06,
+      "loss": 1.1175,
+      "step": 220
+    },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.6422849181469871,
-      "eval_loss": 0.9593217372894287,
-      "eval_runtime": 24.866,
-      "eval_samples_per_second": 115.459,
-      "eval_steps_per_second": 3.619,
-      "step": 202
+      "eval_accuracy": 0.6415436054611312,
+      "eval_loss": 0.9517616629600525,
+      "eval_runtime": 77.1312,
+      "eval_samples_per_second": 93.062,
+      "eval_steps_per_second": 2.917,
+      "step": 224
     },
     {
       "epoch": 1.0,
-      "step": 202,
-      "total_flos": 2.0014719749914337e+18,
-      "train_loss": 1.316133418885788,
-      "train_runtime": 615.4563,
-      "train_samples_per_second": 41.982,
-      "train_steps_per_second": 0.328
+      "step": 224,
+      "total_flos": 2.2210002502885048e+18,
+      "train_loss": 1.2996089096580232,
+      "train_runtime": 698.1963,
+      "train_samples_per_second": 41.119,
+      "train_steps_per_second": 0.321
     }
   ],
-  "max_steps": 202,
+  "max_steps": 224,
   "num_train_epochs": 1,
-  "total_flos": 2.0014719749914337e+18,
+  "total_flos": 2.2210002502885048e+18,
   "trial_name": null,
   "trial_params": null
 }