dino-base-2023_12_18-kornia_img-size518_batch-size32_epochs20

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

• Loss: 0.1422
• F1 Micro: 0.7786
• F1 Macro: 0.7231
• Roc Auc: 0.8542
• Accuracy: 0.4586
• Learning Rate: 0.01

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: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Micro	F1 Macro	Roc Auc	Accuracy	Rate
No log	1.0	268	0.1853	0.6692	0.5725	0.7700	0.4084	0.01
0.2132	2.0	536	0.1678	0.7630	0.7066	0.8729	0.4134	0.01
0.2132	3.0	804	0.1568	0.7665	0.7008	0.8456	0.4541	0.01
0.1913	4.0	1072	0.1496	0.7900	0.7401	0.8786	0.4509	0.01
0.1913	5.0	1340	0.1668	0.7692	0.7355	0.8765	0.4091	0.01
0.1899	6.0	1608	0.1519	0.7619	0.6472	0.8352	0.4605	0.01
0.1899	7.0	1876	0.1590	0.7725	0.6881	0.8654	0.4391	0.01
0.188	8.0	2144	0.1490	0.7812	0.6946	0.8642	0.4459	0.01
0.188	9.0	2412	0.1493	0.7887	0.7115	0.8765	0.4670	0.01
0.1888	10.0	2680	0.1444	0.7744	0.7014	0.8445	0.4720	0.01
0.1888	11.0	2948	0.1582	0.7652	0.6895	0.8498	0.4348	0.01
0.1888	12.0	3216	0.1536	0.7491	0.6946	0.8176	0.4616	0.01
0.1888	13.0	3484	0.1514	0.7728	0.6920	0.8503	0.4555	0.01
0.1886	14.0	3752	0.1668	0.6863	0.5593	0.7725	0.4355	0.01
0.1906	15.0	4020	0.1524	0.7589	0.6660	0.8395	0.4534	0.01
0.1906	16.0	4288	0.1429	0.7849	0.7240	0.8546	0.4762	0.01
0.1879	17.0	4556	0.1711	0.7453	0.6093	0.8492	0.4123	0.01
0.1879	18.0	4824	0.1588	0.7304	0.5857	0.8062	0.4373	0.01
0.1888	19.0	5092	0.1634	0.7428	0.6950	0.8220	0.4466	0.01
0.1888	20.0	5360	0.1485	0.7788	0.7272	0.8566	0.4645	0.01

Framework versions

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