medsiglip-448-ft-vindr

This model is a fine-tuned version of google/medsiglip-448 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.9991

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• 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
• lr_scheduler_warmup_ratio: 0.03
• lr_scheduler_warmup_steps: 300
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss
5.0838	0.4938	100	4.0413
2.9516	0.9877	200	3.9161
2.9046	1.4790	300	4.0662
2.8929	1.9728	400	4.0390
2.8176	2.4642	500	3.9496
2.7783	2.9580	600	3.9971
2.6967	3.4494	700	4.0185
2.6829	3.9432	800	3.9991

Framework versions

• Transformers 4.55.0
• Pytorch 2.8.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4