🍻 cheers

README.md

@@ -1,7 +1,11 @@
 ---
+language:
+- eng
 license: apache-2.0
 base_model: facebook/dinov2-giant
 tags:
+- multilabel-image-classification
+- multilabel
 - generated_from_trainer
 metrics:
 - accuracy
@@ -15,13 +19,13 @@ should probably proofread and complete it, then remove this comment. -->
 
 # dinov2-giant-2024_01_02-kornia_img-size518_batch-size32_epochs20_freeze
 
-This model is a fine-tuned version of [facebook/dinov2-giant](https://huggingface.co/facebook/dinov2-giant) on the None dataset.
+This model is a fine-tuned version of [facebook/dinov2-giant](https://huggingface.co/facebook/dinov2-giant) on the multilabel_complete_dataset dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.1276
-- F1 Micro: 0.8134
-- F1 Macro: 0.7677
-- Roc Auc: 0.8759
-- Accuracy: 0.5263
+- Loss: 0.1266
+- F1 Micro: 0.8142
+- F1 Macro: 0.7719
+- Roc Auc: 0.8801
+- Accuracy: 0.5121
 - Learning Rate: 0.001
 
 ## Model description

all_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 20.0,
+    "eval_accuracy": 0.5121436114044351,
+    "eval_f1_macro": 0.7719123466949233,
+    "eval_f1_micro": 0.8141891891891891,
+    "eval_loss": 0.12663225829601288,
+    "eval_roc_auc": 0.8801205728506106,
+    "eval_runtime": 1345.3407,
+    "eval_samples_per_second": 2.112,
+    "eval_steps_per_second": 0.066,
+    "learning_rate": 0.001,
+    "train_loss": 0.07645123254007367,
+    "train_runtime": 44768.9701,
+    "train_samples_per_second": 3.828,
+    "train_steps_per_second": 0.12
+}

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "facebook/dinov2-giant",
+  "_name_or_path": "facebook/dinov2-giant2024_01_03",
   "apply_layernorm": true,
   "architectures": [
     "NewheadDinov2ForImageClassification"

eval_results.json

@@ -0,0 +1,12 @@
+{
+    "epoch": 20.0,
+    "eval_accuracy": 0.5121436114044351,
+    "eval_f1_macro": 0.7719123466949233,
+    "eval_f1_micro": 0.8141891891891891,
+    "eval_loss": 0.12663225829601288,
+    "eval_roc_auc": 0.8801205728506106,
+    "eval_runtime": 1345.3407,
+    "eval_samples_per_second": 2.112,
+    "eval_steps_per_second": 0.066,
+    "learning_rate": 0.001
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 20.0,
+    "learning_rate": 0.001,
+    "train_loss": 0.07645123254007367,
+    "train_runtime": 44768.9701,
+    "train_samples_per_second": 3.828,
+    "train_steps_per_second": 0.12
+}

trainer_state.json

@@ -0,0 +1,349 @@
+{
+  "best_metric": 0.1276090443134308,
+  "best_model_checkpoint": "/home1/datawork/mcontini/models/multilabel/huggingface/dinov2-giant-2024_01_02-kornia_img-size518_batch-size32_epochs20_freeze/checkpoint-5360",
+  "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": 0.4055019649874955,
+      "eval_f1_macro": 0.5114163498644834,
+      "eval_f1_micro": 0.6258493771234429,
+      "eval_loss": 0.22310702502727509,
+      "eval_roc_auc": 0.7462668565240048,
+      "eval_runtime": 1321.1574,
+      "eval_samples_per_second": 2.119,
+      "eval_steps_per_second": 0.067,
+      "learning_rate": 0.01,
+      "step": 268
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.01,
+      "loss": 0.2273,
+      "step": 500
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.38120757413361916,
+      "eval_f1_macro": 0.45109440187622923,
+      "eval_f1_micro": 0.6105779637780503,
+      "eval_loss": 0.25052765011787415,
+      "eval_roc_auc": 0.7359905994056075,
+      "eval_runtime": 1321.0391,
+      "eval_samples_per_second": 2.119,
+      "eval_steps_per_second": 0.067,
+      "learning_rate": 0.01,
+      "step": 536
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.41764916041443373,
+      "eval_f1_macro": 0.6951782991070653,
+      "eval_f1_micro": 0.7531037857011804,
+      "eval_loss": 0.17819169163703918,
+      "eval_roc_auc": 0.8424985182786072,
+      "eval_runtime": 1324.6672,
+      "eval_samples_per_second": 2.113,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 804
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.01,
+      "loss": 0.196,
+      "step": 1000
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.4240800285816363,
+      "eval_f1_macro": 0.6667105102591024,
+      "eval_f1_micro": 0.7645973453961074,
+      "eval_loss": 0.15775558352470398,
+      "eval_roc_auc": 0.8561540860224954,
+      "eval_runtime": 1324.5164,
+      "eval_samples_per_second": 2.113,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 1072
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.3551268310110754,
+      "eval_f1_macro": 0.6462786847563172,
+      "eval_f1_micro": 0.7289891034631567,
+      "eval_loss": 0.1978054791688919,
+      "eval_roc_auc": 0.8616257579280331,
+      "eval_runtime": 1318.8019,
+      "eval_samples_per_second": 2.122,
+      "eval_steps_per_second": 0.067,
+      "learning_rate": 0.01,
+      "step": 1340
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.01,
+      "loss": 0.1916,
+      "step": 1500
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.45480528760271527,
+      "eval_f1_macro": 0.6154722345590861,
+      "eval_f1_micro": 0.7533968253968253,
+      "eval_loss": 0.1570044308900833,
+      "eval_roc_auc": 0.8332111249225472,
+      "eval_runtime": 1322.0736,
+      "eval_samples_per_second": 2.117,
+      "eval_steps_per_second": 0.067,
+      "learning_rate": 0.01,
+      "step": 1608
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.4076455877098964,
+      "eval_f1_macro": 0.7033703129441511,
+      "eval_f1_micro": 0.7711244893571275,
+      "eval_loss": 0.17040091753005981,
+      "eval_roc_auc": 0.8893350315039463,
+      "eval_runtime": 1321.072,
+      "eval_samples_per_second": 2.119,
+      "eval_steps_per_second": 0.067,
+      "learning_rate": 0.01,
+      "step": 1876
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.01,
+      "loss": 0.1935,
+      "step": 2000
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.44873168988924617,
+      "eval_f1_macro": 0.7240467456541528,
+      "eval_f1_micro": 0.7782660807031561,
+      "eval_loss": 0.1583939492702484,
+      "eval_roc_auc": 0.87591640182751,
+      "eval_runtime": 1324.3647,
+      "eval_samples_per_second": 2.113,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 2144
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.44337263308324404,
+      "eval_f1_macro": 0.7026377863872897,
+      "eval_f1_micro": 0.7724530831099197,
+      "eval_loss": 0.16136306524276733,
+      "eval_roc_auc": 0.878715414884403,
+      "eval_runtime": 1334.0015,
+      "eval_samples_per_second": 2.098,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 2412
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.01,
+      "loss": 0.1945,
+      "step": 2500
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.43658449446230796,
+      "eval_f1_macro": 0.6244582432023046,
+      "eval_f1_micro": 0.743831546311846,
+      "eval_loss": 0.1569277048110962,
+      "eval_roc_auc": 0.8239477455598103,
+      "eval_runtime": 1341.7823,
+      "eval_samples_per_second": 2.086,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 2680
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.42979635584137194,
+      "eval_f1_macro": 0.6986025660751581,
+      "eval_f1_micro": 0.763875932835821,
+      "eval_loss": 0.1665951907634735,
+      "eval_roc_auc": 0.8613857432852049,
+      "eval_runtime": 1333.2352,
+      "eval_samples_per_second": 2.099,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 2948
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.01,
+      "loss": 0.1951,
+      "step": 3000
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.44765987852804573,
+      "eval_f1_macro": 0.6291100964373895,
+      "eval_f1_micro": 0.744811534234176,
+      "eval_loss": 0.15847544372081757,
+      "eval_roc_auc": 0.8242335503635296,
+      "eval_runtime": 1336.9094,
+      "eval_samples_per_second": 2.094,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 3216
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.43801357627724186,
+      "eval_f1_macro": 0.6650314316976844,
+      "eval_f1_micro": 0.7623590633130962,
+      "eval_loss": 0.15646952390670776,
+      "eval_roc_auc": 0.844307766178341,
+      "eval_runtime": 1326.5168,
+      "eval_samples_per_second": 2.11,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 3484
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.01,
+      "loss": 0.1953,
+      "step": 3500
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.4465880671668453,
+      "eval_f1_macro": 0.5021716379213181,
+      "eval_f1_micro": 0.6728278270994033,
+      "eval_loss": 0.17276929318904877,
+      "eval_roc_auc": 0.7638863294586612,
+      "eval_runtime": 1334.5311,
+      "eval_samples_per_second": 2.097,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 3752
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 0.01,
+      "loss": 0.1945,
+      "step": 4000
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.4555198285101822,
+      "eval_f1_macro": 0.6523865470596933,
+      "eval_f1_micro": 0.7441088615997346,
+      "eval_loss": 0.15650007128715515,
+      "eval_roc_auc": 0.8176688310877148,
+      "eval_runtime": 1329.7371,
+      "eval_samples_per_second": 2.105,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 4020
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.4580207216863165,
+      "eval_f1_macro": 0.643940672540041,
+      "eval_f1_micro": 0.7514664626370825,
+      "eval_loss": 0.15759296715259552,
+      "eval_roc_auc": 0.8311410282106092,
+      "eval_runtime": 1337.597,
+      "eval_samples_per_second": 2.093,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 4288
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 0.01,
+      "loss": 0.1929,
+      "step": 4500
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.43122543765630583,
+      "eval_f1_macro": 0.5707187678300992,
+      "eval_f1_micro": 0.7358732093873818,
+      "eval_loss": 0.17005115747451782,
+      "eval_roc_auc": 0.833709678625855,
+      "eval_runtime": 1326.7318,
+      "eval_samples_per_second": 2.11,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 4556
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.42300821722043586,
+      "eval_f1_macro": 0.6533638646137062,
+      "eval_f1_micro": 0.7531492248062015,
+      "eval_loss": 0.15990640223026276,
+      "eval_roc_auc": 0.8450542243393845,
+      "eval_runtime": 1329.7371,
+      "eval_samples_per_second": 2.105,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 4824
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 0.01,
+      "loss": 0.1952,
+      "step": 5000
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.45480528760271527,
+      "eval_f1_macro": 0.6657705432619248,
+      "eval_f1_micro": 0.7347428800106717,
+      "eval_loss": 0.16034173965454102,
+      "eval_roc_auc": 0.8117717147474299,
+      "eval_runtime": 1328.0617,
+      "eval_samples_per_second": 2.108,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.01,
+      "step": 5092
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.5262593783494105,
+      "eval_f1_macro": 0.7676828961065644,
+      "eval_f1_micro": 0.8133643372904173,
+      "eval_loss": 0.1276090443134308,
+      "eval_roc_auc": 0.8759327417838362,
+      "eval_runtime": 1330.6724,
+      "eval_samples_per_second": 2.103,
+      "eval_steps_per_second": 0.066,
+      "learning_rate": 0.001,
+      "step": 5360
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.001,
+      "step": 5360,
+      "total_flos": 9.454976615476292e+20,
+      "train_loss": 0.07645123254007367,
+      "train_runtime": 44768.9701,
+      "train_samples_per_second": 3.828,
+      "train_steps_per_second": 0.12
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 5360,
+  "num_train_epochs": 20,
+  "save_steps": 500,
+  "total_flos": 9.454976615476292e+20,
+  "trial_name": null,
+  "trial_params": null
+}