medsiglip-448-scin-classification

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: 0.3754
• Roc Auc: 0.9760

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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_steps: 5
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Roc Auc
2.8561	0.1985	40	1.3265	0.7629
2.4018	0.3970	80	0.9869	0.8771
1.6941	0.5955	120	0.7613	0.9162
1.4797	0.7940	160	0.6767	0.9398
1.1814	0.9926	200	0.5849	0.9560
0.884	1.1886	240	0.4970	0.9630
0.7607	1.3871	280	0.4639	0.9663
0.7558	1.5856	320	0.4327	0.9694
0.6405	1.7841	360	0.4078	0.9721
0.6299	1.9826	400	0.3851	0.9751
0.4243	2.1787	440	0.3819	0.9753
0.3755	2.3772	480	0.3830	0.9745
0.3719	2.5757	520	0.3793	0.9756
0.36	2.7742	560	0.3756	0.9761
0.3645	2.9727	600	0.3754	0.9760

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1