medsiglip-448-ft-crc100k

This model is a fine-tuned version of google/medsiglip-448 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.2987

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: 0.0001
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Use adamw_torch 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss
1.3955	1.3889	50	0.6563
0.6403	2.7778	100	0.6268
0.6135	4.1667	150	0.6006
0.5883	5.5556	200	0.5760
0.5643	6.9444	250	0.5525
0.5414	8.3333	300	0.5303
0.5198	9.7222	350	0.5092
0.4993	11.1111	400	0.4892
0.4799	12.5	450	0.4705
0.4617	13.8889	500	0.4529
0.4447	15.2778	550	0.4364
0.4288	16.6667	600	0.4211
0.4141	18.0556	650	0.4069
0.4004	19.4444	700	0.3938
0.3878	20.8333	750	0.3818
0.3763	22.2222	800	0.3708
0.3658	23.6111	850	0.3608
0.3562	25.0	900	0.3517
0.3476	26.3889	950	0.3435
0.3399	27.7778	1000	0.3362
0.333	29.1667	1050	0.3297
0.3269	30.5556	1100	0.3241
0.3216	31.9444	1150	0.3191
0.3169	33.3333	1200	0.3148
0.313	34.7222	1250	0.3111
0.3096	36.1111	1300	0.3081
0.3068	37.5	1350	0.3056
0.3045	38.8889	1400	0.3035
0.3027	40.2778	1450	0.3019
0.3013	41.6667	1500	0.3007
0.3003	43.0556	1550	0.2999
0.2996	44.4444	1600	0.2993
0.2991	45.8333	1650	0.2990
0.2988	47.2222	1700	0.2988
0.2987	48.6111	1750	0.2987
0.2987	50.0	1800	0.2987

Framework versions

• Transformers 4.50.3
• Pytorch 2.6.0+cu124
• Datasets 4.5.0
• Tokenizers 0.21.1