retinal_disease_model

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

• Loss: 0.8147
• Accuracy: 0.6062

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.5345	1.0	293	1.3042	0.3991
1.2525	2.0	586	1.3418	0.3633
1.0851	3.0	879	1.2063	0.5008
0.9413	4.0	1172	1.0855	0.4971
0.8212	5.0	1465	0.9673	0.4848
0.7118	6.0	1758	0.9683	0.5350
0.6451	7.0	2051	0.8728	0.6009
0.5910	8.0	2344	0.8177	0.5939
0.5518	9.0	2637	0.8230	0.5993
0.5383	10.0	2930	0.8147	0.6062

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2