End of training

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 5.0,
-    "total_flos": 1.9921933677868186e+18,
-    "train_loss": 0.18438913027445475,
-    "train_runtime": 4974.1556,
-    "train_samples_per_second": 80.416,
-    "train_steps_per_second": 0.106
+    "total_flos": 3.9844365410304e+18,
+    "train_loss": 0.024235197260048972,
+    "train_runtime": 18042.9659,
+    "train_samples_per_second": 44.339,
+    "train_steps_per_second": 0.058
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 5.0,
-    "total_flos": 1.9921933677868186e+18,
-    "train_loss": 0.18438913027445475,
-    "train_runtime": 4974.1556,
-    "train_samples_per_second": 80.416,
-    "train_steps_per_second": 0.106
+    "total_flos": 3.9844365410304e+18,
+    "train_loss": 0.024235197260048972,
+    "train_runtime": 18042.9659,
+    "train_samples_per_second": 44.339,
+    "train_steps_per_second": 0.058
 }

trainer_state.json

@@ -1,132 +1,332 @@
 {
-  "best_metric": 0.028903182595968246,
-  "best_model_checkpoint": "./dinov2_Liveness_detection_v2.2.2/checkpoint-512",
+  "best_metric": 0.021851040422916412,
+  "best_model_checkpoint": "./dinov2_Liveness_detection_v2.2.2/checkpoint-960",
   "epoch": 5.0,
-  "eval_steps": 128,
-  "global_step": 525,
+  "eval_steps": 64,
+  "global_step": 1045,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.6095238095238096,
-      "grad_norm": 31.397323608398438,
-      "learning_rate": 4.428571428571428e-05,
-      "loss": 0.8952,
+      "epoch": 0.3062200956937799,
+      "grad_norm": 16.814943313598633,
+      "learning_rate": 4.6985645933014356e-05,
+      "loss": 0.069,
       "step": 64
     },
     {
-      "epoch": 1.2190476190476192,
-      "grad_norm": 32.01212692260742,
-      "learning_rate": 3.8190476190476196e-05,
-      "loss": 0.266,
+      "epoch": 0.3062200956937799,
+      "eval_accuracy": 0.97885,
+      "eval_f1": 0.9789501971935628,
+      "eval_loss": 0.05791417136788368,
+      "eval_precision": 0.9790213884696916,
+      "eval_recall": 0.97885,
+      "eval_runtime": 521.502,
+      "eval_samples_per_second": 76.702,
+      "eval_steps_per_second": 9.588,
+      "step": 64
+    },
+    {
+      "epoch": 0.6124401913875598,
+      "grad_norm": 11.042802810668945,
+      "learning_rate": 4.3923444976076554e-05,
+      "loss": 0.0482,
       "step": 128
     },
     {
-      "epoch": 1.2190476190476192,
-      "eval_accuracy": 0.942,
-      "eval_f1": 0.9424224282926646,
-      "eval_loss": 0.1499282419681549,
-      "eval_precision": 0.9420692134831461,
-      "eval_recall": 0.942,
-      "eval_runtime": 220.5479,
-      "eval_samples_per_second": 90.683,
-      "eval_steps_per_second": 11.335,
+      "epoch": 0.6124401913875598,
+      "eval_accuracy": 0.981375,
+      "eval_f1": 0.9814562696359047,
+      "eval_loss": 0.05181511119008064,
+      "eval_precision": 0.9815262980731044,
+      "eval_recall": 0.981375,
+      "eval_runtime": 517.0571,
+      "eval_samples_per_second": 77.361,
+      "eval_steps_per_second": 9.67,
       "step": 128
     },
     {
-      "epoch": 1.8285714285714287,
-      "grad_norm": 29.405466079711914,
-      "learning_rate": 3.2095238095238095e-05,
-      "loss": 0.1497,
+      "epoch": 0.9186602870813397,
+      "grad_norm": 5.2479634284973145,
+      "learning_rate": 4.086124401913876e-05,
+      "loss": 0.0404,
+      "step": 192
+    },
+    {
+      "epoch": 0.9186602870813397,
+      "eval_accuracy": 0.9882,
+      "eval_f1": 0.9881867637379899,
+      "eval_loss": 0.032511744648218155,
+      "eval_precision": 0.9882451896307357,
+      "eval_recall": 0.9882,
+      "eval_runtime": 490.7248,
+      "eval_samples_per_second": 81.512,
+      "eval_steps_per_second": 10.189,
       "step": 192
     },
     {
-      "epoch": 2.4380952380952383,
-      "grad_norm": 8.088194847106934,
-      "learning_rate": 2.6000000000000002e-05,
-      "loss": 0.0929,
+      "epoch": 1.2248803827751196,
+      "grad_norm": 3.8472037315368652,
+      "learning_rate": 3.7799043062200956e-05,
+      "loss": 0.0325,
       "step": 256
     },
     {
-      "epoch": 2.4380952380952383,
-      "eval_accuracy": 0.95735,
-      "eval_f1": 0.9578785031500121,
-      "eval_loss": 0.12672634422779083,
-      "eval_precision": 0.9593868981160767,
-      "eval_recall": 0.95735,
-      "eval_runtime": 209.2644,
-      "eval_samples_per_second": 95.573,
-      "eval_steps_per_second": 11.947,
+      "epoch": 1.2248803827751196,
+      "eval_accuracy": 0.98545,
+      "eval_f1": 0.985494798685368,
+      "eval_loss": 0.04072735831141472,
+      "eval_precision": 0.9855132158274779,
+      "eval_recall": 0.98545,
+      "eval_runtime": 466.7395,
+      "eval_samples_per_second": 85.701,
+      "eval_steps_per_second": 10.713,
       "step": 256
     },
     {
-      "epoch": 3.0476190476190474,
-      "grad_norm": 27.274545669555664,
-      "learning_rate": 1.9904761904761908e-05,
-      "loss": 0.0533,
+      "epoch": 1.5311004784688995,
+      "grad_norm": 4.665475368499756,
+      "learning_rate": 3.473684210526316e-05,
+      "loss": 0.0322,
       "step": 320
     },
     {
-      "epoch": 3.657142857142857,
-      "grad_norm": 7.902669906616211,
-      "learning_rate": 1.3809523809523811e-05,
-      "loss": 0.0298,
+      "epoch": 1.5311004784688995,
+      "eval_accuracy": 0.99015,
+      "eval_f1": 0.990145076314261,
+      "eval_loss": 0.028076525777578354,
+      "eval_precision": 0.9901636180839065,
+      "eval_recall": 0.99015,
+      "eval_runtime": 480.6976,
+      "eval_samples_per_second": 83.212,
+      "eval_steps_per_second": 10.402,
+      "step": 320
+    },
+    {
+      "epoch": 1.8373205741626795,
+      "grad_norm": 5.757445335388184,
+      "learning_rate": 3.1674641148325365e-05,
+      "loss": 0.0283,
       "step": 384
     },
     {
-      "epoch": 3.657142857142857,
-      "eval_accuracy": 0.9888,
-      "eval_f1": 0.9887893352142478,
-      "eval_loss": 0.030348431318998337,
-      "eval_precision": 0.9888352116245671,
-      "eval_recall": 0.9888,
-      "eval_runtime": 205.3865,
-      "eval_samples_per_second": 97.377,
-      "eval_steps_per_second": 12.172,
+      "epoch": 1.8373205741626795,
+      "eval_accuracy": 0.988425,
+      "eval_f1": 0.9884000225035877,
+      "eval_loss": 0.0346578024327755,
+      "eval_precision": 0.9885429088726858,
+      "eval_recall": 0.988425,
+      "eval_runtime": 483.3989,
+      "eval_samples_per_second": 82.747,
+      "eval_steps_per_second": 10.343,
       "step": 384
     },
     {
-      "epoch": 4.266666666666667,
-      "grad_norm": 4.974146366119385,
-      "learning_rate": 7.714285714285714e-06,
-      "loss": 0.0159,
+      "epoch": 2.1435406698564594,
+      "grad_norm": 2.869988203048706,
+      "learning_rate": 2.8612440191387562e-05,
+      "loss": 0.0256,
+      "step": 448
+    },
+    {
+      "epoch": 2.1435406698564594,
+      "eval_accuracy": 0.990725,
+      "eval_f1": 0.9907227766051799,
+      "eval_loss": 0.027070404961705208,
+      "eval_precision": 0.9907303189959034,
+      "eval_recall": 0.990725,
+      "eval_runtime": 480.102,
+      "eval_samples_per_second": 83.316,
+      "eval_steps_per_second": 10.414,
       "step": 448
     },
     {
-      "epoch": 4.876190476190477,
-      "grad_norm": 9.068910598754883,
-      "learning_rate": 1.619047619047619e-06,
-      "loss": 0.0083,
+      "epoch": 2.449760765550239,
+      "grad_norm": 4.530367374420166,
+      "learning_rate": 2.5550239234449763e-05,
+      "loss": 0.0207,
       "step": 512
     },
     {
-      "epoch": 4.876190476190477,
+      "epoch": 2.449760765550239,
+      "eval_accuracy": 0.987375,
+      "eval_f1": 0.9873992902183978,
+      "eval_loss": 0.035906460136175156,
+      "eval_precision": 0.9873848219691154,
+      "eval_recall": 0.987375,
+      "eval_runtime": 485.3186,
+      "eval_samples_per_second": 82.42,
+      "eval_steps_per_second": 10.303,
+      "step": 512
+    },
+    {
+      "epoch": 2.7559808612440193,
+      "grad_norm": 2.7303965091705322,
+      "learning_rate": 2.2488038277511964e-05,
+      "loss": 0.0192,
+      "step": 576
+    },
+    {
+      "epoch": 2.7559808612440193,
+      "eval_accuracy": 0.99165,
+      "eval_f1": 0.9916437240635789,
+      "eval_loss": 0.025301363319158554,
+      "eval_precision": 0.9916710521622637,
+      "eval_recall": 0.99165,
+      "eval_runtime": 481.2714,
+      "eval_samples_per_second": 83.113,
+      "eval_steps_per_second": 10.389,
+      "step": 576
+    },
+    {
+      "epoch": 3.062200956937799,
+      "grad_norm": 2.5271918773651123,
+      "learning_rate": 1.9425837320574165e-05,
+      "loss": 0.017,
+      "step": 640
+    },
+    {
+      "epoch": 3.062200956937799,
+      "eval_accuracy": 0.9908,
+      "eval_f1": 0.9908068152778651,
+      "eval_loss": 0.02719452604651451,
+      "eval_precision": 0.9907948609341658,
+      "eval_recall": 0.9908,
+      "eval_runtime": 486.6359,
+      "eval_samples_per_second": 82.197,
+      "eval_steps_per_second": 10.275,
+      "step": 640
+    },
+    {
+      "epoch": 3.3684210526315788,
+      "grad_norm": 3.4637880325317383,
+      "learning_rate": 1.6363636363636366e-05,
+      "loss": 0.0134,
+      "step": 704
+    },
+    {
+      "epoch": 3.3684210526315788,
       "eval_accuracy": 0.99155,
-      "eval_f1": 0.9915498405407145,
-      "eval_loss": 0.028903182595968246,
-      "eval_precision": 0.9915503244910029,
+      "eval_f1": 0.991553221477779,
+      "eval_loss": 0.025475142523646355,
+      "eval_precision": 0.9915457922828868,
       "eval_recall": 0.99155,
-      "eval_runtime": 209.0101,
-      "eval_samples_per_second": 95.689,
-      "eval_steps_per_second": 11.961,
-      "step": 512
+      "eval_runtime": 487.284,
+      "eval_samples_per_second": 82.088,
+      "eval_steps_per_second": 10.261,
+      "step": 704
+    },
+    {
+      "epoch": 3.674641148325359,
+      "grad_norm": 2.8516900539398193,
+      "learning_rate": 1.3301435406698565e-05,
+      "loss": 0.0132,
+      "step": 768
+    },
+    {
+      "epoch": 3.674641148325359,
+      "eval_accuracy": 0.9925,
+      "eval_f1": 0.9924949448411151,
+      "eval_loss": 0.023235656321048737,
+      "eval_precision": 0.9925169638113596,
+      "eval_recall": 0.9925,
+      "eval_runtime": 489.1056,
+      "eval_samples_per_second": 81.782,
+      "eval_steps_per_second": 10.223,
+      "step": 768
+    },
+    {
+      "epoch": 3.9808612440191387,
+      "grad_norm": 2.9499707221984863,
+      "learning_rate": 1.0239234449760766e-05,
+      "loss": 0.0114,
+      "step": 832
+    },
+    {
+      "epoch": 3.9808612440191387,
+      "eval_accuracy": 0.99185,
+      "eval_f1": 0.9918568051307793,
+      "eval_loss": 0.026025714352726936,
+      "eval_precision": 0.9918472125621515,
+      "eval_recall": 0.99185,
+      "eval_runtime": 486.1176,
+      "eval_samples_per_second": 82.285,
+      "eval_steps_per_second": 10.286,
+      "step": 832
+    },
+    {
+      "epoch": 4.287081339712919,
+      "grad_norm": 3.4067459106445312,
+      "learning_rate": 7.177033492822967e-06,
+      "loss": 0.0074,
+      "step": 896
+    },
+    {
+      "epoch": 4.287081339712919,
+      "eval_accuracy": 0.992725,
+      "eval_f1": 0.9927214337347136,
+      "eval_loss": 0.024225549772381783,
+      "eval_precision": 0.9927357715365919,
+      "eval_recall": 0.992725,
+      "eval_runtime": 482.0799,
+      "eval_samples_per_second": 82.974,
+      "eval_steps_per_second": 10.372,
+      "step": 896
+    },
+    {
+      "epoch": 4.5933014354066986,
+      "grad_norm": 3.6422884464263916,
+      "learning_rate": 4.114832535885168e-06,
+      "loss": 0.0079,
+      "step": 960
+    },
+    {
+      "epoch": 4.5933014354066986,
+      "eval_accuracy": 0.99315,
+      "eval_f1": 0.9931459136485375,
+      "eval_loss": 0.021851040422916412,
+      "eval_precision": 0.9931635536599013,
+      "eval_recall": 0.99315,
+      "eval_runtime": 488.3316,
+      "eval_samples_per_second": 81.912,
+      "eval_steps_per_second": 10.239,
+      "step": 960
+    },
+    {
+      "epoch": 4.899521531100478,
+      "grad_norm": 4.330822467803955,
+      "learning_rate": 1.0526315789473685e-06,
+      "loss": 0.0072,
+      "step": 1024
+    },
+    {
+      "epoch": 4.899521531100478,
+      "eval_accuracy": 0.993425,
+      "eval_f1": 0.9934216498623175,
+      "eval_loss": 0.02193431928753853,
+      "eval_precision": 0.9934356337892436,
+      "eval_recall": 0.993425,
+      "eval_runtime": 492.599,
+      "eval_samples_per_second": 81.202,
+      "eval_steps_per_second": 10.15,
+      "step": 1024
     },
     {
       "epoch": 5.0,
-      "step": 525,
-      "total_flos": 1.9921933677868186e+18,
-      "train_loss": 0.18438913027445475,
-      "train_runtime": 4974.1556,
-      "train_samples_per_second": 80.416,
-      "train_steps_per_second": 0.106
+      "step": 1045,
+      "total_flos": 3.9844365410304e+18,
+      "train_loss": 0.024235197260048972,
+      "train_runtime": 18042.9659,
+      "train_samples_per_second": 44.339,
+      "train_steps_per_second": 0.058
     }
   ],
   "logging_steps": 64,
-  "max_steps": 525,
+  "max_steps": 1045,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 5,
-  "save_steps": 128,
+  "save_steps": 64,
   "stateful_callbacks": {
     "TrainerControl": {
       "args": {
@@ -139,7 +339,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.9921933677868186e+18,
+  "total_flos": 3.9844365410304e+18,
   "train_batch_size": 768,
   "trial_name": null,
   "trial_params": null