🍻 cheers

README.md

@@ -1,7 +1,11 @@
 ---
+language:
+- eng
 license: apache-2.0
 base_model: facebook/dinov2-base
 tags:
+- multilabel-image-classification
+- multilabel
 - generated_from_trainer
 metrics:
 - accuracy
@@ -15,13 +19,13 @@ should probably proofread and complete it, then remove this comment. -->
 
 # dino-base-2023_12_18-kornia_img-size518_batch-size32_epochs20
 
-This model is a fine-tuned version of [facebook/dinov2-base](https://huggingface.co/facebook/dinov2-base) on the None dataset.
+This model is a fine-tuned version of [facebook/dinov2-base](https://huggingface.co/facebook/dinov2-base) on the multilabel_complete_dataset dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1485
-- F1 Micro: 0.7788
-- F1 Macro: 0.7272
-- Roc Auc: 0.8566
-- Accuracy: 0.4645
+- Loss: 0.1422
+- F1 Micro: 0.7786
+- F1 Macro: 0.7231
+- Roc Auc: 0.8542
+- Accuracy: 0.4586
 - Learning Rate: 0.01
 
 ## Model description

all_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 20.0,
+    "eval_accuracy": 0.4586413234776487,
+    "eval_f1_macro": 0.7230670728714794,
+    "eval_f1_micro": 0.7785516178736519,
+    "eval_loss": 0.14224284887313843,
+    "eval_roc_auc": 0.854225641152639,
+    "eval_runtime": 453.9537,
+    "eval_samples_per_second": 6.258,
+    "eval_steps_per_second": 0.196,
+    "learning_rate": 0.01,
+    "train_loss": 0.19166865989343443,
+    "train_runtime": 38664.1493,
+    "train_samples_per_second": 4.432,
+    "train_steps_per_second": 0.139
+}

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "facebook/dinov2-base",
+  "_name_or_path": "facebook/dinov2-base2023_12_18",
   "apply_layernorm": true,
   "architectures": [
     "NewheadDinov2ForImageClassification"

eval_results.json

@@ -0,0 +1,12 @@
+{
+    "epoch": 20.0,
+    "eval_accuracy": 0.4586413234776487,
+    "eval_f1_macro": 0.7230670728714794,
+    "eval_f1_micro": 0.7785516178736519,
+    "eval_loss": 0.14224284887313843,
+    "eval_roc_auc": 0.854225641152639,
+    "eval_runtime": 453.9537,
+    "eval_samples_per_second": 6.258,
+    "eval_steps_per_second": 0.196,
+    "learning_rate": 0.01
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 20.0,
+    "learning_rate": 0.01,
+    "train_loss": 0.19166865989343443,
+    "train_runtime": 38664.1493,
+    "train_samples_per_second": 4.432,
+    "train_steps_per_second": 0.139
+}

trainer_state.json

@@ -0,0 +1,349 @@
+{
+  "best_metric": 0.1429145485162735,
+  "best_model_checkpoint": "/home1/datawork/mcontini/models/multilabel/huggingface/dino-base-2023_12_18-kornia_img-size518_batch-size32_epochs20/checkpoint-4288",
+  "epoch": 20.0,
+  "eval_steps": 500,
+  "global_step": 5360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.40836012861736337,
+      "eval_f1_macro": 0.5724832983501338,
+      "eval_f1_micro": 0.66917502787068,
+      "eval_loss": 0.18527662754058838,
+      "eval_roc_auc": 0.7700155485850371,
+      "eval_runtime": 452.1812,
+      "eval_samples_per_second": 6.19,
+      "eval_steps_per_second": 0.195,
+      "learning_rate": 0.01,
+      "step": 268
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.01,
+      "loss": 0.2132,
+      "step": 500
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.41336191496963204,
+      "eval_f1_macro": 0.7066155100478595,
+      "eval_f1_micro": 0.7630167597765364,
+      "eval_loss": 0.1677507609128952,
+      "eval_roc_auc": 0.8728909820573381,
+      "eval_runtime": 452.3557,
+      "eval_samples_per_second": 6.188,
+      "eval_steps_per_second": 0.195,
+      "learning_rate": 0.01,
+      "step": 536
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.4540907466952483,
+      "eval_f1_macro": 0.7007992492814997,
+      "eval_f1_micro": 0.7664760009947775,
+      "eval_loss": 0.15684403479099274,
+      "eval_roc_auc": 0.8456356297999449,
+      "eval_runtime": 449.5285,
+      "eval_samples_per_second": 6.227,
+      "eval_steps_per_second": 0.196,
+      "learning_rate": 0.01,
+      "step": 804
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.01,
+      "loss": 0.1913,
+      "step": 1000
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.45087531261164704,
+      "eval_f1_macro": 0.7401496915878392,
+      "eval_f1_micro": 0.7900162149640955,
+      "eval_loss": 0.14960218966007233,
+      "eval_roc_auc": 0.8786412916916798,
+      "eval_runtime": 447.0415,
+      "eval_samples_per_second": 6.261,
+      "eval_steps_per_second": 0.197,
+      "learning_rate": 0.01,
+      "step": 1072
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.4090746695248303,
+      "eval_f1_macro": 0.7354739664497006,
+      "eval_f1_micro": 0.7692135710692527,
+      "eval_loss": 0.16684336960315704,
+      "eval_roc_auc": 0.8765264813882132,
+      "eval_runtime": 449.4387,
+      "eval_samples_per_second": 6.228,
+      "eval_steps_per_second": 0.196,
+      "learning_rate": 0.01,
+      "step": 1340
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.01,
+      "loss": 0.1899,
+      "step": 1500
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.46052161486245086,
+      "eval_f1_macro": 0.6472064428456584,
+      "eval_f1_micro": 0.7618986529826812,
+      "eval_loss": 0.1519184708595276,
+      "eval_roc_auc": 0.8352141402262778,
+      "eval_runtime": 452.3131,
+      "eval_samples_per_second": 6.188,
+      "eval_steps_per_second": 0.195,
+      "learning_rate": 0.01,
+      "step": 1608
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.4390853876384423,
+      "eval_f1_macro": 0.6880839155457914,
+      "eval_f1_micro": 0.7724533380141596,
+      "eval_loss": 0.15900948643684387,
+      "eval_roc_auc": 0.8654028748541202,
+      "eval_runtime": 449.1274,
+      "eval_samples_per_second": 6.232,
+      "eval_steps_per_second": 0.196,
+      "learning_rate": 0.01,
+      "step": 1876
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.01,
+      "loss": 0.188,
+      "step": 2000
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.44587352625937837,
+      "eval_f1_macro": 0.6946270568821779,
+      "eval_f1_micro": 0.7812088569718731,
+      "eval_loss": 0.14899832010269165,
+      "eval_roc_auc": 0.8641522922073566,
+      "eval_runtime": 449.6916,
+      "eval_samples_per_second": 6.224,
+      "eval_steps_per_second": 0.196,
+      "learning_rate": 0.01,
+      "step": 2144
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.4669524830296534,
+      "eval_f1_macro": 0.7114660813932495,
+      "eval_f1_micro": 0.7886651925986268,
+      "eval_loss": 0.14934837818145752,
+      "eval_roc_auc": 0.8764578257293321,
+      "eval_runtime": 456.6917,
+      "eval_samples_per_second": 6.129,
+      "eval_steps_per_second": 0.193,
+      "learning_rate": 0.01,
+      "step": 2412
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.01,
+      "loss": 0.1888,
+      "step": 2500
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.47195426938192214,
+      "eval_f1_macro": 0.7013528531179296,
+      "eval_f1_micro": 0.7743778567800914,
+      "eval_loss": 0.14436914026737213,
+      "eval_roc_auc": 0.8445106252570078,
+      "eval_runtime": 457.228,
+      "eval_samples_per_second": 6.122,
+      "eval_steps_per_second": 0.192,
+      "learning_rate": 0.01,
+      "step": 2680
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.4347981421936406,
+      "eval_f1_macro": 0.6894667259492441,
+      "eval_f1_micro": 0.7651653850848285,
+      "eval_loss": 0.158202663064003,
+      "eval_roc_auc": 0.8497870587564459,
+      "eval_runtime": 451.0784,
+      "eval_samples_per_second": 6.205,
+      "eval_steps_per_second": 0.195,
+      "learning_rate": 0.01,
+      "step": 2948
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.01,
+      "loss": 0.1888,
+      "step": 3000
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.4615934262236513,
+      "eval_f1_macro": 0.6945655735071663,
+      "eval_f1_micro": 0.7490938380990738,
+      "eval_loss": 0.15359686315059662,
+      "eval_roc_auc": 0.8176277103144143,
+      "eval_runtime": 456.0533,
+      "eval_samples_per_second": 6.137,
+      "eval_steps_per_second": 0.193,
+      "learning_rate": 0.01,
+      "step": 3216
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.4555198285101822,
+      "eval_f1_macro": 0.691978470964451,
+      "eval_f1_micro": 0.7727919787089188,
+      "eval_loss": 0.15135115385055542,
+      "eval_roc_auc": 0.8502953769902751,
+      "eval_runtime": 449.1842,
+      "eval_samples_per_second": 6.231,
+      "eval_steps_per_second": 0.196,
+      "learning_rate": 0.01,
+      "step": 3484
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.01,
+      "loss": 0.1886,
+      "step": 3500
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.4355126831011075,
+      "eval_f1_macro": 0.5592534839805278,
+      "eval_f1_micro": 0.686313973548016,
+      "eval_loss": 0.16679102182388306,
+      "eval_roc_auc": 0.7724894168127243,
+      "eval_runtime": 450.5535,
+      "eval_samples_per_second": 6.212,
+      "eval_steps_per_second": 0.195,
+      "learning_rate": 0.01,
+      "step": 3752
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 0.01,
+      "loss": 0.1906,
+      "step": 4000
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.4533762057877814,
+      "eval_f1_macro": 0.6660464840620275,
+      "eval_f1_micro": 0.7589447960398522,
+      "eval_loss": 0.15235908329486847,
+      "eval_roc_auc": 0.8395164037430609,
+      "eval_runtime": 453.9816,
+      "eval_samples_per_second": 6.165,
+      "eval_steps_per_second": 0.194,
+      "learning_rate": 0.01,
+      "step": 4020
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.47624151482672383,
+      "eval_f1_macro": 0.7239567261315109,
+      "eval_f1_micro": 0.784867024597328,
+      "eval_loss": 0.1429145485162735,
+      "eval_roc_auc": 0.8546121997155703,
+      "eval_runtime": 451.7346,
+      "eval_samples_per_second": 6.196,
+      "eval_steps_per_second": 0.195,
+      "learning_rate": 0.01,
+      "step": 4288
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 0.01,
+      "loss": 0.1879,
+      "step": 4500
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.4122901036084316,
+      "eval_f1_macro": 0.6093139734586531,
+      "eval_f1_micro": 0.7452841241944933,
+      "eval_loss": 0.17112578451633453,
+      "eval_roc_auc": 0.8492466394980158,
+      "eval_runtime": 451.5598,
+      "eval_samples_per_second": 6.199,
+      "eval_steps_per_second": 0.195,
+      "learning_rate": 0.01,
+      "step": 4556
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.43729903536977494,
+      "eval_f1_macro": 0.5856617023433228,
+      "eval_f1_micro": 0.7304124060404957,
+      "eval_loss": 0.15878728032112122,
+      "eval_roc_auc": 0.8061813000254404,
+      "eval_runtime": 453.038,
+      "eval_samples_per_second": 6.178,
+      "eval_steps_per_second": 0.194,
+      "learning_rate": 0.01,
+      "step": 4824
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 0.01,
+      "loss": 0.1888,
+      "step": 5000
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.4465880671668453,
+      "eval_f1_macro": 0.69497597215645,
+      "eval_f1_micro": 0.7428348605966075,
+      "eval_loss": 0.16344697773456573,
+      "eval_roc_auc": 0.821961971607228,
+      "eval_runtime": 451.5696,
+      "eval_samples_per_second": 6.198,
+      "eval_steps_per_second": 0.195,
+      "learning_rate": 0.01,
+      "step": 5092
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.4644515898535191,
+      "eval_f1_macro": 0.7272077268448706,
+      "eval_f1_micro": 0.7788184879093969,
+      "eval_loss": 0.1484629064798355,
+      "eval_roc_auc": 0.856568523463588,
+      "eval_runtime": 450.0153,
+      "eval_samples_per_second": 6.22,
+      "eval_steps_per_second": 0.196,
+      "learning_rate": 0.01,
+      "step": 5360
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.01,
+      "step": 5360,
+      "total_flos": 7.288742611336668e+19,
+      "train_loss": 0.19166865989343443,
+      "train_runtime": 38664.1493,
+      "train_samples_per_second": 4.432,
+      "train_steps_per_second": 0.139
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 5360,
+  "num_train_epochs": 20,
+  "save_steps": 500,
+  "total_flos": 7.288742611336668e+19,
+  "trial_name": null,
+  "trial_params": null
+}