vit-base-patch16-224-new-finetuned-ecg-classification-google-vit

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

• Loss: 0.1945
• Accuracy: 0.9462

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: 0.0005
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.8889	6	1.2109	0.4839
1.2698	1.9259	13	2.1752	0.3011
1.03	2.9630	20	1.0502	0.5591
1.03	4.0	27	1.0719	0.5914
0.7666	4.8889	33	0.6510	0.7419
0.6305	5.9259	40	0.4610	0.8065
0.6305	6.9630	47	0.6421	0.6667
0.6328	8.0	54	0.4628	0.8172
0.4884	8.8889	60	0.3666	0.8817
0.4884	9.9259	67	0.6400	0.7419
0.4872	10.9630	74	0.4497	0.8172
0.3887	12.0	81	0.3410	0.8602
0.3887	12.8889	87	0.3811	0.8602
0.3002	13.9259	94	0.1945	0.9462
0.3012	14.9630	101	0.2167	0.9140
0.3012	16.0	108	0.2265	0.9355
0.3045	16.8889	114	0.1877	0.9140
0.2534	17.7778	120	0.1712	0.9462

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1