dino-large-2023_12_18-kornia_img-size518_batch-size32_epochs20

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: 0.1139
• F1 Micro: 0.8312
• F1 Macro: 0.7939
• Roc Auc: 0.8886
• Accuracy: 0.5276
• 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	0.1825	0.6774	0.5984	0.7768	0.4237	0.01
0.2147	2.0	536	0.1749	0.7567	0.6636	0.8530	0.4309	0.01
0.2147	3.0	804	0.1555	0.7735	0.7210	0.8586	0.4469	0.01
0.1902	4.0	1072	0.1582	0.7546	0.6775	0.8368	0.4469	0.01
0.1902	5.0	1340	0.1599	0.7757	0.7218	0.8741	0.4323	0.01
0.1908	6.0	1608	0.1546	0.7598	0.6469	0.8373	0.4712	0.01
0.1908	7.0	1876	0.1502	0.7747	0.7194	0.8592	0.4566	0.01
0.1871	8.0	2144	0.1542	0.7759	0.7180	0.8669	0.4416	0.01
0.1871	9.0	2412	0.1494	0.7853	0.7134	0.8666	0.4812	0.01
0.1906	10.0	2680	0.1515	0.7530	0.6504	0.8264	0.4673	0.01
0.1906	11.0	2948	0.1578	0.7495	0.6769	0.8296	0.4387	0.01
0.1897	12.0	3216	0.1557	0.7757	0.6902	0.8594	0.4387	0.01
0.1897	13.0	3484	0.1612	0.7293	0.6498	0.8054	0.4530	0.01
0.1879	14.0	3752	0.1800	0.6633	0.5069	0.7571	0.4577	0.01
0.1902	15.0	4020	0.1674	0.7432	0.6672	0.8349	0.4541	0.01
0.1902	16.0	4288	0.1287	0.8137	0.7725	0.8764	0.5177	0.001
0.1612	17.0	4556	0.1246	0.8286	0.7943	0.8927	0.5227	0.001
0.1612	18.0	4824	0.1235	0.8271	0.7885	0.8856	0.5384	0.001
0.1479	19.0	5092	0.1203	0.8366	0.8018	0.8941	0.5423	0.001
0.1479	20.0	5360	0.1148	0.8330	0.7934	0.8852	0.5384	0.001

Framework versions

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