ViT_breastmnist

This model is a fine-tuned version of google/vit-base-patch16-224 on the medmnist-v2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3570
• Accuracy: 0.8654
• F1: 0.8157

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: 5e-05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.5391	0.5556	10	0.4007	0.7949	0.6698
0.3685	1.1111	20	0.3650	0.8718	0.8120
0.2275	1.6667	30	0.3601	0.8462	0.8101
0.1604	2.2222	40	0.2938	0.8718	0.8319
0.0624	2.7778	50	0.2966	0.8846	0.8511
0.0597	3.3333	60	0.4313	0.8974	0.8556
0.029	3.8889	70	0.4105	0.8718	0.8194
0.0094	4.4444	80	0.3746	0.9103	0.8803
0.0077	5.0	90	0.4098	0.8974	0.8655
0.0082	5.5556	100	0.4451	0.9103	0.8803
0.0024	6.1111	110	0.4599	0.8974	0.8655
0.0028	6.6667	120	0.4739	0.8974	0.8608
0.0013	7.2222	130	0.4653	0.8974	0.8655
0.0016	7.7778	140	0.4927	0.8974	0.8608
0.0011	8.3333	150	0.5115	0.8974	0.8608
0.0015	8.8889	160	0.5055	0.8974	0.8608
0.0007	9.4444	170	0.4982	0.8974	0.8608
0.0011	10.0	180	0.4975	0.8974	0.8608

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0