dino-large-2023_12_15-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: 0.1159
• F1 Micro: 0.8297
• F1 Macro: 0.7800
• Roc Auc: 0.8880
• Accuracy: 0.5435
• 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: 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	0.1749	0.6911	0.5979	0.7869	0.4355	0.01
0.2108	2.0	536	0.1887	0.6448	0.5545	0.7473	0.4330	0.01
0.2108	3.0	804	0.1639	0.7291	0.5960	0.8089	0.4223	0.01
0.1888	4.0	1072	0.1676	0.7426	0.6583	0.8357	0.4416	0.01
0.1888	5.0	1340	0.1619	0.7133	0.6050	0.7932	0.4527	0.01
0.1877	6.0	1608	0.1532	0.7497	0.6624	0.8279	0.4691	0.01
0.1877	7.0	1876	0.1690	0.7409	0.6522	0.8402	0.3758	0.01
0.1845	8.0	2144	0.1461	0.7830	0.6987	0.8601	0.4770	0.01
0.1845	9.0	2412	0.1533	0.7629	0.7143	0.8392	0.4659	0.01
0.1845	10.0	2680	0.1703	0.7335	0.6227	0.8306	0.4234	0.01
0.1845	11.0	2948	0.1714	0.7191	0.6140	0.8054	0.4455	0.01
0.1871	12.0	3216	0.1538	0.7677	0.6816	0.8427	0.4562	0.01
0.1871	13.0	3484	0.1533	0.7673	0.6923	0.8508	0.4573	0.01
0.1881	14.0	3752	0.1572	0.7550	0.6424	0.8395	0.4584	0.01
0.1768	15.0	4020	0.1300	0.8118	0.7646	0.8766	0.5248	0.001
0.1768	16.0	4288	0.1237	0.8229	0.7727	0.8838	0.5277	0.001
0.1499	17.0	4556	0.1218	0.8285	0.7891	0.8933	0.5348	0.001
0.1499	18.0	4824	0.1193	0.8285	0.7831	0.8902	0.5284	0.001
0.1424	19.0	5092	0.1182	0.8298	0.7778	0.8849	0.5445	0.001
0.1424	20.0	5360	0.1182	0.8286	0.7698	0.8827	0.5395	0.001

Framework versions

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