dino-large-2023_12_06-with_custom_head

This model is a fine-tuned version of facebook/dinov2-large on the multilabel_complete_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.2014
• F1 Micro: 0.8291
• F1 Macro: 0.8015
• Roc Auc: 0.9029
• Accuracy: 0.5132
• Learning Rate: 0.001

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.01
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 14

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Micro	F1 Macro	Roc Auc	Accuracy	Rate
0.4706	1.0	536	0.4533	0.7389	0.6876	0.8316	0.4269	0.01
0.4045	2.0	1072	0.4262	0.7669	0.7188	0.8634	0.4391	0.01
0.3973	3.0	1608	0.4722	0.7601	0.7176	0.8372	0.4537	0.01
0.3961	4.0	2144	0.6075	0.7528	0.6913	0.8724	0.3762	0.01
0.3751	5.0	2680	0.3916	0.7884	0.7511	0.8925	0.4352	0.01
0.365	6.0	3216	0.5256	0.7660	0.7066	0.8535	0.4105	0.01
0.3565	7.0	3752	0.5708	0.7293	0.6947	0.8254	0.4101	0.01
0.3807	8.0	4288	0.4770	0.7811	0.7145	0.8609	0.4591	0.01
0.3462	9.0	4824	0.4612	0.7880	0.7522	0.8775	0.4452	0.01
0.38	10.0	5360	0.4559	0.7943	0.7517	0.8747	0.4612	0.01
0.3472	11.0	5896	0.5081	0.7709	0.7315	0.8980	0.4041	0.01
0.3167	12.0	6432	0.2364	0.8268	0.7990	0.8945	0.5141	0.001
0.1322	13.0	6968	0.2222	0.8209	0.7931	0.8951	0.4923	0.001
0.0958	14.0	7504	0.2089	0.8287	0.7975	0.8985	0.5052	0.001

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1