vit_itri_downsample_normal

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

• Loss: 6.6738
• Accuracy: 0.3355
• Precision: 0.4688
• Recall: 0.3355
• F1: 0.2701

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: 24
• eval_batch_size: 4
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.3133	1.0	334	3.3993	0.3685	0.5578	0.3685	0.3520
0.0981	2.0	668	4.3016	0.3603	0.5583	0.3603	0.3196
0.0628	3.0	1002	4.9479	0.3386	0.3274	0.3386	0.2681
0.0466	4.0	1336	4.8740	0.3150	0.4302	0.3150	0.2640
0.0297	5.0	1670	6.3499	0.3087	0.2861	0.3087	0.2053
0.0195	6.0	2004	5.5555	0.3498	0.4355	0.3498	0.2700
0.0261	7.0	2338	5.7446	0.3446	0.4763	0.3446	0.2888
0.0186	8.0	2672	6.0125	0.3107	0.3748	0.3107	0.2341
0.0117	9.0	3006	5.8823	0.3099	0.3911	0.3099	0.2456
0.0116	10.0	3340	5.9882	0.3331	0.4533	0.3331	0.2682
0.0048	11.0	3674	5.7636	0.3028	0.4634	0.3028	0.2980
0.0095	12.0	4008	6.1077	0.3228	0.4431	0.3228	0.2770
0.0011	13.0	4342	6.2826	0.3135	0.4474	0.3135	0.2744
0.0021	14.0	4676	6.2547	0.3503	0.4589	0.3503	0.2997
0.0013	15.0	5010	6.2053	0.3472	0.4804	0.3472	0.3102
0.0009	16.0	5344	6.6696	0.3405	0.4513	0.3405	0.2825
0.0001	17.0	5678	6.8090	0.3467	0.4797	0.3467	0.2853
0.0015	18.0	6012	6.5591	0.3379	0.4661	0.3379	0.2813
0.0001	19.0	6346	6.6722	0.3345	0.4674	0.3345	0.2724
0.0004	20.0	6680	6.6738	0.3355	0.4688	0.3355	0.2701

Framework versions

• Transformers 4.53.0.dev0
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1