vit-base-aihub_model-v2

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.3076
• Accuracy: 0.9639
• Precision: 0.9610
• Recall: 0.9614
• F1: 0.9604

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: 128
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 512
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
No log	1.0	3	1.2753	0.8373	0.8563	0.7993	0.8022
No log	2.0	6	1.1252	0.8675	0.8895	0.8300	0.8333
No log	3.0	9	0.9427	0.8976	0.9185	0.8696	0.8760
1.1721	4.0	12	0.7995	0.9398	0.9474	0.9195	0.9246
1.1721	5.0	15	0.6820	0.9699	0.9704	0.9613	0.9642
1.1721	6.0	18	0.5927	0.9639	0.9603	0.9583	0.9587
0.7084	7.0	21	0.5239	0.9759	0.9725	0.9729	0.9725
0.7084	8.0	24	0.4743	0.9699	0.9665	0.9671	0.9665
0.7084	9.0	27	0.4436	0.9578	0.9558	0.9556	0.9544
0.4668	10.0	30	0.4070	0.9639	0.9610	0.9614	0.9604
0.4668	11.0	33	0.3817	0.9699	0.9665	0.9671	0.9665
0.4668	12.0	36	0.3625	0.9699	0.9665	0.9671	0.9665
0.4668	13.0	39	0.3536	0.9578	0.9558	0.9556	0.9544
0.3611	14.0	42	0.3384	0.9578	0.9558	0.9556	0.9544
0.3611	15.0	45	0.3249	0.9699	0.9665	0.9671	0.9665
0.3611	16.0	48	0.3164	0.9699	0.9665	0.9671	0.9665
0.3063	17.0	51	0.3142	0.9639	0.9610	0.9614	0.9604
0.3063	18.0	54	0.3122	0.9639	0.9610	0.9614	0.9604
0.3063	19.0	57	0.3093	0.9639	0.9610	0.9614	0.9604
0.294	20.0	60	0.3076	0.9639	0.9610	0.9614	0.9604

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu117
• Datasets 2.12.0
• Tokenizers 0.13.3