Fin de l'entraînement

all_results.json

@@ -1,14 +1,14 @@
 {
     "epoch": 4.0,
-    "eval_loss": 6.522369384765625,
-    "eval_mae": 2.7217423915863037,
-    "eval_mse": 7.407970428466797,
-    "eval_runtime": 2.0059,
-    "eval_samples_per_second": 4.985,
-    "eval_steps_per_second": 0.997,
+    "eval_loss": 0.13025157153606415,
+    "eval_mae": 0.41358014941215515,
+    "eval_mse": 0.21276399493217468,
+    "eval_runtime": 1.9358,
+    "eval_samples_per_second": 5.166,
+    "eval_steps_per_second": 1.033,
     "total_flos": 0.0,
-    "train_loss": 15.188075065612793,
-    "train_runtime": 31.017,
-    "train_samples_per_second": 10.059,
-    "train_steps_per_second": 0.258
+    "train_loss": 0.2500866651535034,
+    "train_runtime": 34.0562,
+    "train_samples_per_second": 9.161,
+    "train_steps_per_second": 0.235
 }

eval_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 4.0,
-    "eval_loss": 6.522369384765625,
-    "eval_mae": 2.7217423915863037,
-    "eval_mse": 7.407970428466797,
-    "eval_runtime": 2.0059,
-    "eval_samples_per_second": 4.985,
-    "eval_steps_per_second": 0.997
+    "eval_loss": 0.13025157153606415,
+    "eval_mae": 0.41358014941215515,
+    "eval_mse": 0.21276399493217468,
+    "eval_runtime": 1.9358,
+    "eval_samples_per_second": 5.166,
+    "eval_steps_per_second": 1.033
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 4.0,
     "total_flos": 0.0,
-    "train_loss": 15.188075065612793,
-    "train_runtime": 31.017,
-    "train_samples_per_second": 10.059,
-    "train_steps_per_second": 0.258
+    "train_loss": 0.2500866651535034,
+    "train_runtime": 34.0562,
+    "train_samples_per_second": 9.161,
+    "train_steps_per_second": 0.235
 }

trainer_state.json

@@ -1,6 +1,6 @@
 {
   "best_global_step": 8,
-  "best_metric": 7.407970428466797,
+  "best_metric": 0.21276399493217468,
   "best_model_checkpoint": "./vit_focus/checkpoint-8",
   "epoch": 4.0,
   "eval_steps": 500,
@@ -11,52 +11,52 @@
   "log_history": [
     {
       "epoch": 1.0,
-      "eval_loss": 6.890933036804199,
-      "eval_mae": 2.7386960983276367,
-      "eval_mse": 7.500483512878418,
-      "eval_runtime": 1.2828,
-      "eval_samples_per_second": 7.795,
-      "eval_steps_per_second": 1.559,
+      "eval_loss": 0.14130041003227234,
+      "eval_mae": 0.4174715280532837,
+      "eval_mse": 0.21645615994930267,
+      "eval_runtime": 1.4102,
+      "eval_samples_per_second": 7.091,
+      "eval_steps_per_second": 1.418,
       "step": 2
     },
     {
       "epoch": 2.0,
-      "eval_loss": 6.799283504486084,
-      "eval_mae": 2.7345499992370605,
-      "eval_mse": 7.4778008460998535,
-      "eval_runtime": 1.1777,
-      "eval_samples_per_second": 8.491,
-      "eval_steps_per_second": 1.698,
+      "eval_loss": 0.13852722942829132,
+      "eval_mae": 0.41653603315353394,
+      "eval_mse": 0.21554431319236755,
+      "eval_runtime": 1.1496,
+      "eval_samples_per_second": 8.699,
+      "eval_steps_per_second": 1.74,
       "step": 4
     },
     {
       "epoch": 3.0,
-      "eval_loss": 6.6634087562561035,
-      "eval_mae": 2.728336811065674,
-      "eval_mse": 7.443875789642334,
-      "eval_runtime": 1.1717,
-      "eval_samples_per_second": 8.535,
-      "eval_steps_per_second": 1.707,
+      "eval_loss": 0.13479307293891907,
+      "eval_mae": 0.4152373969554901,
+      "eval_mse": 0.2143227458000183,
+      "eval_runtime": 1.1869,
+      "eval_samples_per_second": 8.425,
+      "eval_steps_per_second": 1.685,
       "step": 6
     },
     {
       "epoch": 4.0,
-      "eval_loss": 6.522369384765625,
-      "eval_mae": 2.7217423915863037,
-      "eval_mse": 7.407970428466797,
-      "eval_runtime": 1.2243,
-      "eval_samples_per_second": 8.168,
-      "eval_steps_per_second": 1.634,
+      "eval_loss": 0.13025157153606415,
+      "eval_mae": 0.41358014941215515,
+      "eval_mse": 0.21276399493217468,
+      "eval_runtime": 1.193,
+      "eval_samples_per_second": 8.382,
+      "eval_steps_per_second": 1.676,
       "step": 8
     },
     {
       "epoch": 4.0,
       "step": 8,
       "total_flos": 0.0,
-      "train_loss": 15.188075065612793,
-      "train_runtime": 31.017,
-      "train_samples_per_second": 10.059,
-      "train_steps_per_second": 0.258
+      "train_loss": 0.2500866651535034,
+      "train_runtime": 34.0562,
+      "train_samples_per_second": 9.161,
+      "train_steps_per_second": 0.235
     }
   ],
   "logging_steps": 50,