End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 5.17,
-    "eval_accuracy": 0.8987341772151899,
-    "eval_loss": 0.492713063955307,
-    "eval_runtime": 66.3108,
-    "eval_samples_per_second": 2.383,
-    "eval_steps_per_second": 2.383
+    "eval_accuracy": 0.9117647058823529,
+    "eval_loss": 0.37310147285461426,
+    "eval_runtime": 33.7759,
+    "eval_samples_per_second": 2.013,
+    "eval_steps_per_second": 2.013
 }

test_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 5.17,
-    "eval_accuracy": 0.8987341772151899,
-    "eval_loss": 0.492713063955307,
-    "eval_runtime": 66.3108,
-    "eval_samples_per_second": 2.383,
-    "eval_steps_per_second": 2.383
+    "eval_accuracy": 0.9117647058823529,
+    "eval_loss": 0.37310147285461426,
+    "eval_runtime": 33.7759,
+    "eval_samples_per_second": 2.013,
+    "eval_steps_per_second": 2.013
 }

trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.9285714285714286,
+  "best_metric": 0.9583333333333334,
   "best_model_checkpoint": "timesformer-base-finetuned-k400-finetuned-kinetic-subset-three-local-temporal-with-spatial/checkpoint-150",
   "epoch": 5.166666666666667,
   "eval_steps": 500,
@@ -10,302 +10,293 @@
   "log_history": [
     {
       "epoch": 0.03,
-      "grad_norm": 32.2842903137207,
+      "grad_norm": 0.01975257694721222,
       "learning_rate": 1.6666666666666667e-05,
-      "loss": 1.6309,
+      "loss": 0.0009,
       "step": 10
     },
     {
       "epoch": 0.07,
-      "grad_norm": 29.811689376831055,
+      "grad_norm": 0.014412298798561096,
       "learning_rate": 3.3333333333333335e-05,
-      "loss": 1.7024,
+      "loss": 0.0005,
       "step": 20
     },
     {
       "epoch": 0.1,
-      "grad_norm": 26.413127899169922,
+      "grad_norm": 0.048455920070409775,
       "learning_rate": 5e-05,
-      "loss": 1.5057,
+      "loss": 0.0018,
       "step": 30
     },
     {
       "epoch": 0.13,
-      "grad_norm": 17.640932083129883,
+      "grad_norm": 0.003363919910043478,
       "learning_rate": 4.814814814814815e-05,
-      "loss": 0.8117,
+      "loss": 0.0012,
       "step": 40
     },
     {
       "epoch": 0.17,
-      "grad_norm": 33.26347732543945,
+      "grad_norm": 0.024853387847542763,
       "learning_rate": 4.62962962962963e-05,
-      "loss": 0.2937,
+      "loss": 0.0003,
       "step": 50
     },
     {
       "epoch": 0.17,
-      "eval_accuracy": 0.8690476190476191,
-      "eval_loss": 0.3884676396846771,
-      "eval_runtime": 105.2382,
-      "eval_samples_per_second": 1.596,
-      "eval_steps_per_second": 1.596,
+      "eval_accuracy": 0.8472222222222222,
+      "eval_loss": 0.4552870988845825,
+      "eval_runtime": 58.6213,
+      "eval_samples_per_second": 1.228,
+      "eval_steps_per_second": 1.228,
       "step": 50
     },
     {
       "epoch": 1.03,
-      "grad_norm": 0.8504588007926941,
+      "grad_norm": 0.004102244041860104,
       "learning_rate": 4.4444444444444447e-05,
-      "loss": 0.4217,
+      "loss": 0.3192,
       "step": 60
     },
     {
       "epoch": 1.07,
-      "grad_norm": 0.1911085844039917,
+      "grad_norm": 0.003750218078494072,
       "learning_rate": 4.259259259259259e-05,
-      "loss": 0.2901,
+      "loss": 0.0002,
       "step": 70
     },
     {
       "epoch": 1.1,
-      "grad_norm": 0.12057169526815414,
+      "grad_norm": 0.002901990432292223,
       "learning_rate": 4.074074074074074e-05,
-      "loss": 0.0256,
+      "loss": 0.5754,
       "step": 80
     },
     {
       "epoch": 1.13,
-      "grad_norm": 0.2651258111000061,
+      "grad_norm": 0.019022488966584206,
       "learning_rate": 3.888888888888889e-05,
-      "loss": 0.0627,
+      "loss": 0.0002,
       "step": 90
     },
     {
       "epoch": 1.17,
-      "grad_norm": 0.1569805145263672,
+      "grad_norm": 0.004384249914437532,
       "learning_rate": 3.7037037037037037e-05,
-      "loss": 0.2365,
+      "loss": 0.0004,
       "step": 100
     },
     {
       "epoch": 1.17,
-      "eval_accuracy": 0.8809523809523809,
-      "eval_loss": 0.600234866142273,
-      "eval_runtime": 109.7974,
-      "eval_samples_per_second": 1.53,
-      "eval_steps_per_second": 1.53,
+      "eval_accuracy": 0.9027777777777778,
+      "eval_loss": 0.33492106199264526,
+      "eval_runtime": 58.3208,
+      "eval_samples_per_second": 1.235,
+      "eval_steps_per_second": 1.235,
       "step": 100
     },
     {
       "epoch": 2.03,
-      "grad_norm": 0.038823891431093216,
+      "grad_norm": 0.008162049576640129,
       "learning_rate": 3.518518518518519e-05,
-      "loss": 0.0043,
+      "loss": 0.0003,
       "step": 110
     },
     {
       "epoch": 2.07,
-      "grad_norm": 0.037996720522642136,
+      "grad_norm": 0.0084863705560565,
       "learning_rate": 3.3333333333333335e-05,
-      "loss": 0.0062,
+      "loss": 0.0002,
       "step": 120
     },
     {
       "epoch": 2.1,
-      "grad_norm": 0.05053324997425079,
+      "grad_norm": 0.005603179335594177,
       "learning_rate": 3.148148148148148e-05,
-      "loss": 0.0297,
+      "loss": 0.0004,
       "step": 130
     },
     {
       "epoch": 2.13,
-      "grad_norm": 1.2805042266845703,
+      "grad_norm": 0.035682689398527145,
       "learning_rate": 2.962962962962963e-05,
-      "loss": 0.0046,
+      "loss": 0.0003,
       "step": 140
     },
     {
       "epoch": 2.17,
-      "grad_norm": 0.19171175360679626,
+      "grad_norm": 0.07813193649053574,
       "learning_rate": 2.777777777777778e-05,
-      "loss": 0.0164,
+      "loss": 0.0004,
       "step": 150
     },
     {
       "epoch": 2.17,
-      "eval_accuracy": 0.9285714285714286,
-      "eval_loss": 0.2311260849237442,
-      "eval_runtime": 108.5839,
-      "eval_samples_per_second": 1.547,
-      "eval_steps_per_second": 1.547,
+      "eval_accuracy": 0.9583333333333334,
+      "eval_loss": 0.12790486216545105,
+      "eval_runtime": 58.0641,
+      "eval_samples_per_second": 1.24,
+      "eval_steps_per_second": 1.24,
       "step": 150
     },
     {
       "epoch": 3.03,
-      "grad_norm": 0.025192299857735634,
+      "grad_norm": 0.0019370431546121836,
       "learning_rate": 2.5925925925925925e-05,
-      "loss": 0.1078,
+      "loss": 0.0003,
       "step": 160
     },
     {
       "epoch": 3.07,
-      "grad_norm": 0.02108108066022396,
+      "grad_norm": 0.0020284331403672695,
       "learning_rate": 2.4074074074074074e-05,
-      "loss": 0.6885,
+      "loss": 0.5874,
       "step": 170
     },
     {
       "epoch": 3.1,
-      "grad_norm": 0.008117140270769596,
+      "grad_norm": 0.0013429196551442146,
       "learning_rate": 2.2222222222222223e-05,
-      "loss": 0.0011,
+      "loss": 0.019,
       "step": 180
     },
     {
       "epoch": 3.13,
-      "grad_norm": 0.0680527612566948,
+      "grad_norm": 3.899311065673828,
       "learning_rate": 2.037037037037037e-05,
-      "loss": 0.0015,
+      "loss": 0.0037,
       "step": 190
     },
     {
       "epoch": 3.17,
-      "grad_norm": 0.011410138569772243,
+      "grad_norm": 0.001104432623833418,
       "learning_rate": 1.8518518518518518e-05,
-      "loss": 0.4057,
+      "loss": 0.0995,
       "step": 200
     },
     {
       "epoch": 3.17,
-      "eval_accuracy": 0.8630952380952381,
-      "eval_loss": 0.5045905113220215,
-      "eval_runtime": 109.1813,
-      "eval_samples_per_second": 1.539,
-      "eval_steps_per_second": 1.539,
+      "eval_accuracy": 0.9305555555555556,
+      "eval_loss": 0.1180504709482193,
+      "eval_runtime": 57.5974,
+      "eval_samples_per_second": 1.25,
+      "eval_steps_per_second": 1.25,
       "step": 200
     },
     {
       "epoch": 4.03,
-      "grad_norm": 0.015822872519493103,
+      "grad_norm": 0.003288655774667859,
       "learning_rate": 1.6666666666666667e-05,
-      "loss": 0.0012,
+      "loss": 0.0003,
       "step": 210
     },
     {
       "epoch": 4.07,
-      "grad_norm": 0.06866960972547531,
+      "grad_norm": 0.018307456746697426,
       "learning_rate": 1.4814814814814815e-05,
-      "loss": 0.0008,
+      "loss": 0.0002,
       "step": 220
     },
     {
       "epoch": 4.1,
-      "grad_norm": 0.025955336168408394,
+      "grad_norm": 0.012207794934511185,
       "learning_rate": 1.2962962962962962e-05,
-      "loss": 0.0014,
+      "loss": 0.0004,
       "step": 230
     },
     {
       "epoch": 4.13,
-      "grad_norm": 0.027819635346531868,
+      "grad_norm": 0.001230477704666555,
       "learning_rate": 1.1111111111111112e-05,
-      "loss": 0.034,
+      "loss": 0.0007,
       "step": 240
     },
     {
       "epoch": 4.17,
-      "grad_norm": 0.011835525743663311,
+      "grad_norm": 0.001645821612328291,
       "learning_rate": 9.259259259259259e-06,
-      "loss": 0.0004,
+      "loss": 0.0001,
       "step": 250
     },
     {
       "epoch": 4.17,
-      "eval_accuracy": 0.875,
-      "eval_loss": 0.4492361545562744,
-      "eval_runtime": 108.7199,
-      "eval_samples_per_second": 1.545,
-      "eval_steps_per_second": 1.545,
+      "eval_accuracy": 0.9166666666666666,
+      "eval_loss": 0.17973226308822632,
+      "eval_runtime": 57.4339,
+      "eval_samples_per_second": 1.254,
+      "eval_steps_per_second": 1.254,
       "step": 250
     },
     {
       "epoch": 5.03,
-      "grad_norm": 0.9948074817657471,
+      "grad_norm": 0.04015364125370979,
       "learning_rate": 7.4074074074074075e-06,
-      "loss": 0.0022,
+      "loss": 0.0002,
       "step": 260
     },
     {
       "epoch": 5.07,
-      "grad_norm": 0.07322775572538376,
+      "grad_norm": 0.10828184336423874,
       "learning_rate": 5.555555555555556e-06,
-      "loss": 0.0011,
+      "loss": 0.0002,
       "step": 270
     },
     {
       "epoch": 5.1,
-      "grad_norm": 0.01132030040025711,
+      "grad_norm": 0.0032853526063263416,
       "learning_rate": 3.7037037037037037e-06,
-      "loss": 0.0146,
+      "loss": 0.0006,
       "step": 280
     },
     {
       "epoch": 5.13,
-      "grad_norm": 48.047698974609375,
+      "grad_norm": 0.005327914375811815,
       "learning_rate": 1.8518518518518519e-06,
       "loss": 0.0825,
       "step": 290
     },
     {
       "epoch": 5.17,
-      "grad_norm": 0.02245236746966839,
+      "grad_norm": 0.004417418502271175,
       "learning_rate": 0.0,
-      "loss": 0.0007,
+      "loss": 0.0003,
       "step": 300
     },
     {
       "epoch": 5.17,
-      "eval_accuracy": 0.8928571428571429,
-      "eval_loss": 0.35668858885765076,
-      "eval_runtime": 108.8318,
-      "eval_samples_per_second": 1.544,
-      "eval_steps_per_second": 1.544,
+      "eval_accuracy": 0.9444444444444444,
+      "eval_loss": 0.13933943212032318,
+      "eval_runtime": 56.7608,
+      "eval_samples_per_second": 1.268,
+      "eval_steps_per_second": 1.268,
       "step": 300
     },
     {
       "epoch": 5.17,
       "step": 300,
       "total_flos": 3.28565653420032e+17,
-      "train_loss": 0.2795241907471791,
-      "train_runtime": 838.6117,
-      "train_samples_per_second": 0.358,
-      "train_steps_per_second": 0.358
-    },
-    {
-      "epoch": 5.17,
-      "eval_accuracy": 0.8987341772151899,
-      "eval_loss": 0.49271300435066223,
-      "eval_runtime": 66.5373,
-      "eval_samples_per_second": 2.375,
-      "eval_steps_per_second": 2.375,
-      "step": 300
+      "train_loss": 0.056568971110973504,
+      "train_runtime": 562.6502,
+      "train_samples_per_second": 0.533,
+      "train_steps_per_second": 0.533
     },
     {
       "epoch": 5.17,
-      "eval_accuracy": 0.8987341772151899,
-      "eval_loss": 0.492713063955307,
-      "eval_runtime": 65.6092,
-      "eval_samples_per_second": 2.408,
-      "eval_steps_per_second": 2.408,
+      "eval_accuracy": 0.9117647058823529,
+      "eval_loss": 0.37310144305229187,
+      "eval_runtime": 34.9776,
+      "eval_samples_per_second": 1.944,
+      "eval_steps_per_second": 1.944,
       "step": 300
     },
     {
       "epoch": 5.17,
-      "eval_accuracy": 0.8987341772151899,
-      "eval_loss": 0.492713063955307,
-      "eval_runtime": 66.3108,
-      "eval_samples_per_second": 2.383,
-      "eval_steps_per_second": 2.383,
+      "eval_accuracy": 0.9117647058823529,
+      "eval_loss": 0.37310147285461426,
+      "eval_runtime": 33.7759,
+      "eval_samples_per_second": 2.013,
+      "eval_steps_per_second": 2.013,
       "step": 300
     }
   ],