🍻 cheers

README.md

@@ -1,7 +1,11 @@
 ---
+language:
+- eng
 license: apache-2.0
 base_model: facebook/dinov2-large
 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-large-2023_12_06-with_custom_head
 
-This model is a fine-tuned version of [facebook/dinov2-large](https://huggingface.co/facebook/dinov2-large) on the None dataset.
+This model is a fine-tuned version of [facebook/dinov2-large](https://huggingface.co/facebook/dinov2-large) on the multilabel_complete_dataset dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.2089
-- F1 Micro: 0.8287
-- F1 Macro: 0.7975
-- Roc Auc: 0.8985
-- Accuracy: 0.5052
+- Loss: 0.2014
+- F1 Micro: 0.8291
+- F1 Macro: 0.8015
+- Roc Auc: 0.9029
+- Accuracy: 0.5132
 - Learning Rate: 0.001
 
 ## Model description

all_results.json

@@ -1,17 +1,16 @@
 {
-    "epoch": 1.0,
-    "eval_accuracy": 0.43470608940513905,
-    "eval_f1_macro": 0.7133257665684506,
-    "eval_f1_micro": 0.7480296154764748,
-    "eval_loss": 0.39359405636787415,
-    "eval_roc_auc": 0.845573070259888,
-    "eval_runtime": 675.0063,
-    "eval_samples_per_second": 4.209,
-    "eval_steps_per_second": 0.264,
-    "learning_rate": 0.01,
-    "total_flos": 2.3556420369950147e+18,
-    "train_loss": 0.4620438476107014,
-    "train_runtime": 2703.0478,
-    "train_samples_per_second": 3.17,
-    "train_steps_per_second": 0.198
+    "epoch": 14.0,
+    "eval_accuracy": 0.5131995776135164,
+    "eval_f1_macro": 0.8014693809171014,
+    "eval_f1_micro": 0.8291474654377881,
+    "eval_loss": 0.20140419900417328,
+    "eval_roc_auc": 0.9029293443336694,
+    "eval_runtime": 708.7743,
+    "eval_samples_per_second": 4.008,
+    "eval_steps_per_second": 0.251,
+    "learning_rate": 0.001,
+    "train_loss": 0.32461663065498064,
+    "train_runtime": 40410.0145,
+    "train_samples_per_second": 2.968,
+    "train_steps_per_second": 0.186
 }

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "facebook/dinov2-large",
+  "_name_or_path": "facebook/dinov2-large2023_12_06",
   "apply_layernorm": true,
   "architectures": [
     "NewheadDinov2ForImageClassification"

eval_results.json

@@ -1,12 +1,12 @@
 {
-    "epoch": 1.0,
-    "eval_accuracy": 0.43470608940513905,
-    "eval_f1_macro": 0.7133257665684506,
-    "eval_f1_micro": 0.7480296154764748,
-    "eval_loss": 0.39359405636787415,
-    "eval_roc_auc": 0.845573070259888,
-    "eval_runtime": 675.0063,
-    "eval_samples_per_second": 4.209,
-    "eval_steps_per_second": 0.264,
-    "learning_rate": 0.01
+    "epoch": 14.0,
+    "eval_accuracy": 0.5131995776135164,
+    "eval_f1_macro": 0.8014693809171014,
+    "eval_f1_micro": 0.8291474654377881,
+    "eval_loss": 0.20140419900417328,
+    "eval_roc_auc": 0.9029293443336694,
+    "eval_runtime": 708.7743,
+    "eval_samples_per_second": 4.008,
+    "eval_steps_per_second": 0.251,
+    "learning_rate": 0.001
 }

train_results.json

@@ -1,9 +1,8 @@
 {
-    "epoch": 1.0,
-    "learning_rate": 0.01,
-    "total_flos": 2.3556420369950147e+18,
-    "train_loss": 0.4620438476107014,
-    "train_runtime": 2703.0478,
-    "train_samples_per_second": 3.17,
-    "train_steps_per_second": 0.198
+    "epoch": 14.0,
+    "learning_rate": 0.001,
+    "train_loss": 0.32461663065498064,
+    "train_runtime": 40410.0145,
+    "train_samples_per_second": 2.968,
+    "train_steps_per_second": 0.186
 }

trainer_state.json

@@ -1,9 +1,9 @@
 {
-  "best_metric": 0.3991844356060028,
-  "best_model_checkpoint": "/home1/datahome/mcontini/multilabelTest/huggingface_multilabel/models/dino-large-2023_12_06-with_custom_head/checkpoint-536",
-  "epoch": 1.0,
+  "best_metric": 0.20891834795475006,
+  "best_model_checkpoint": "/home1/datahome/mcontini/multilabelTest/huggingface_multilabel/models/dino-large-2023_12_06-with_custom_head/checkpoint-7504",
+  "epoch": 14.0,
   "eval_steps": 500,
-  "global_step": 536,
+  "global_step": 7504,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -11,38 +11,291 @@
     {
       "epoch": 0.93,
       "learning_rate": 0.01,
-      "loss": 0.4649,
+      "loss": 0.4706,
       "step": 500
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.43265451947123973,
-      "eval_f1_macro": 0.7143014999669115,
-      "eval_f1_micro": 0.7494419306184013,
-      "eval_loss": 0.3991844356060028,
-      "eval_roc_auc": 0.8439493091840017,
-      "eval_runtime": 667.7282,
-      "eval_samples_per_second": 4.192,
-      "eval_steps_per_second": 0.262,
+      "eval_accuracy": 0.4269381922115041,
+      "eval_f1_macro": 0.6875675206337933,
+      "eval_f1_micro": 0.7388882010647516,
+      "eval_loss": 0.45332154631614685,
+      "eval_roc_auc": 0.8315732046841607,
+      "eval_runtime": 721.2005,
+      "eval_samples_per_second": 3.881,
+      "eval_steps_per_second": 0.243,
       "learning_rate": 0.01,
       "step": 536
     },
     {
-      "epoch": 1.0,
+      "epoch": 1.87,
+      "learning_rate": 0.01,
+      "loss": 0.4045,
+      "step": 1000
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.4390853876384423,
+      "eval_f1_macro": 0.7188299260461222,
+      "eval_f1_micro": 0.7669365643385565,
+      "eval_loss": 0.42623651027679443,
+      "eval_roc_auc": 0.8634183746872706,
+      "eval_runtime": 732.1356,
+      "eval_samples_per_second": 3.823,
+      "eval_steps_per_second": 0.239,
+      "learning_rate": 0.01,
+      "step": 1072
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.01,
+      "loss": 0.3973,
+      "step": 1500
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.45373347624151483,
+      "eval_f1_macro": 0.7176495048915421,
+      "eval_f1_micro": 0.7600811564798376,
+      "eval_loss": 0.47221142053604126,
+      "eval_roc_auc": 0.8371515138013037,
+      "eval_runtime": 722.5503,
+      "eval_samples_per_second": 3.874,
+      "eval_steps_per_second": 0.242,
+      "learning_rate": 0.01,
+      "step": 1608
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.01,
+      "loss": 0.3961,
+      "step": 2000
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.3762057877813505,
+      "eval_f1_macro": 0.6912769640128906,
+      "eval_f1_micro": 0.7527505610597187,
+      "eval_loss": 0.6075000166893005,
+      "eval_roc_auc": 0.8724199945622111,
+      "eval_runtime": 721.5253,
+      "eval_samples_per_second": 3.879,
+      "eval_steps_per_second": 0.243,
+      "learning_rate": 0.01,
+      "step": 2144
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.01,
+      "loss": 0.3751,
+      "step": 2500
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.43515541264737406,
+      "eval_f1_macro": 0.7510721393725562,
+      "eval_f1_micro": 0.7884265637916782,
+      "eval_loss": 0.3916389048099518,
+      "eval_roc_auc": 0.8925436873239415,
+      "eval_runtime": 719.4104,
+      "eval_samples_per_second": 3.891,
+      "eval_steps_per_second": 0.243,
+      "learning_rate": 0.01,
+      "step": 2680
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.01,
+      "loss": 0.365,
+      "step": 3000
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.4105037513397642,
+      "eval_f1_macro": 0.7065890629165198,
+      "eval_f1_micro": 0.7659574468085106,
+      "eval_loss": 0.5255631804466248,
+      "eval_roc_auc": 0.8535409911580616,
+      "eval_runtime": 733.8873,
+      "eval_samples_per_second": 3.814,
+      "eval_steps_per_second": 0.238,
+      "learning_rate": 0.01,
+      "step": 3216
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.01,
+      "loss": 0.3565,
+      "step": 3500
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.41014648088603073,
+      "eval_f1_macro": 0.6946929982635759,
+      "eval_f1_micro": 0.7292831407628276,
+      "eval_loss": 0.5707747936248779,
+      "eval_roc_auc": 0.8254485110934348,
+      "eval_runtime": 708.4289,
+      "eval_samples_per_second": 3.951,
+      "eval_steps_per_second": 0.247,
+      "learning_rate": 0.01,
+      "step": 3752
+    },
+    {
+      "epoch": 7.46,
       "learning_rate": 0.01,
-      "step": 536,
-      "total_flos": 2.3556420369950147e+18,
-      "train_loss": 0.4620438476107014,
-      "train_runtime": 2703.0478,
-      "train_samples_per_second": 3.17,
-      "train_steps_per_second": 0.198
+      "loss": 0.3807,
+      "step": 4000
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.45909253304751696,
+      "eval_f1_macro": 0.7145318802135835,
+      "eval_f1_micro": 0.7811138014527845,
+      "eval_loss": 0.47700363397598267,
+      "eval_roc_auc": 0.8609308266040633,
+      "eval_runtime": 693.0017,
+      "eval_samples_per_second": 4.039,
+      "eval_steps_per_second": 0.253,
+      "learning_rate": 0.01,
+      "step": 4288
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.01,
+      "loss": 0.3462,
+      "step": 4500
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.4451589853519114,
+      "eval_f1_macro": 0.7521525226987047,
+      "eval_f1_micro": 0.7880472550382212,
+      "eval_loss": 0.4611993134021759,
+      "eval_roc_auc": 0.8774820735887874,
+      "eval_runtime": 688.9161,
+      "eval_samples_per_second": 4.063,
+      "eval_steps_per_second": 0.254,
+      "learning_rate": 0.01,
+      "step": 4824
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.01,
+      "loss": 0.38,
+      "step": 5000
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.46123615576991783,
+      "eval_f1_macro": 0.7516573553144021,
+      "eval_f1_micro": 0.7942691374104554,
+      "eval_loss": 0.4558601677417755,
+      "eval_roc_auc": 0.8746914793507129,
+      "eval_runtime": 694.0203,
+      "eval_samples_per_second": 4.033,
+      "eval_steps_per_second": 0.252,
+      "learning_rate": 0.01,
+      "step": 5360
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.01,
+      "loss": 0.3472,
+      "step": 5500
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.4040728831725616,
+      "eval_f1_macro": 0.7314807716558429,
+      "eval_f1_micro": 0.7708779443254817,
+      "eval_loss": 0.5080511569976807,
+      "eval_roc_auc": 0.8980409880768458,
+      "eval_runtime": 691.0636,
+      "eval_samples_per_second": 4.05,
+      "eval_steps_per_second": 0.253,
+      "learning_rate": 0.01,
+      "step": 5896
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.001,
+      "loss": 0.3167,
+      "step": 6000
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.5141121829224723,
+      "eval_f1_macro": 0.7990159125560728,
+      "eval_f1_micro": 0.826813005251667,
+      "eval_loss": 0.2364267259836197,
+      "eval_roc_auc": 0.8944540329754681,
+      "eval_runtime": 711.5081,
+      "eval_samples_per_second": 3.934,
+      "eval_steps_per_second": 0.246,
+      "learning_rate": 0.001,
+      "step": 6432
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.001,
+      "loss": 0.1322,
+      "step": 6500
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.49231868524473027,
+      "eval_f1_macro": 0.793143079384716,
+      "eval_f1_micro": 0.8209398474349269,
+      "eval_loss": 0.22222588956356049,
+      "eval_roc_auc": 0.8951348180852069,
+      "eval_runtime": 708.4317,
+      "eval_samples_per_second": 3.951,
+      "eval_steps_per_second": 0.247,
+      "learning_rate": 0.001,
+      "step": 6968
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.001,
+      "loss": 0.1074,
+      "step": 7000
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 0.001,
+      "loss": 0.0958,
+      "step": 7500
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.5051804215791355,
+      "eval_f1_macro": 0.797475400821107,
+      "eval_f1_micro": 0.8286666276962646,
+      "eval_loss": 0.20891834795475006,
+      "eval_roc_auc": 0.8984516760867104,
+      "eval_runtime": 704.2208,
+      "eval_samples_per_second": 3.975,
+      "eval_steps_per_second": 0.249,
+      "learning_rate": 0.001,
+      "step": 7504
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 0.001,
+      "step": 7504,
+      "total_flos": 3.2978988517930205e+19,
+      "train_loss": 0.32461663065498064,
+      "train_runtime": 40410.0145,
+      "train_samples_per_second": 2.968,
+      "train_steps_per_second": 0.186
     }
   ],
   "logging_steps": 500,
-  "max_steps": 536,
-  "num_train_epochs": 1,
+  "max_steps": 7504,
+  "num_train_epochs": 14,
   "save_steps": 500,
-  "total_flos": 2.3556420369950147e+18,
+  "total_flos": 3.2978988517930205e+19,
   "trial_name": null,
   "trial_params": null
 }