vit-base-patch16-224-ve-U13-b-24

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

• Loss: 0.5896
• Accuracy: 0.8478

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: 5.5e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 24

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.92	6	1.3800	0.4565
1.3792	2.0	13	1.3093	0.5870
1.3792	2.92	19	1.2228	0.5
1.2786	4.0	26	1.1303	0.5652
1.1265	4.92	32	1.0615	0.5435
1.1265	6.0	39	1.0205	0.4565
0.9906	6.92	45	0.9259	0.6304
0.8632	8.0	52	0.8739	0.7391
0.8632	8.92	58	0.8381	0.7609
0.7529	10.0	65	0.7604	0.7826
0.6468	10.92	71	0.7212	0.8043
0.6468	12.0	78	0.6825	0.7826
0.5553	12.92	84	0.6409	0.8261
0.4704	14.0	91	0.6471	0.8261
0.4704	14.92	97	0.6296	0.7609
0.415	16.0	104	0.5896	0.8478
0.3444	16.92	110	0.5828	0.8043
0.3444	18.0	117	0.5771	0.8261
0.3212	18.92	123	0.5672	0.8261
0.3021	20.0	130	0.5596	0.8478
0.3021	20.92	136	0.5527	0.8261
0.3004	22.0	143	0.5429	0.8261
0.3004	22.15	144	0.5427	0.8261

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu118
• Datasets 2.16.1
• Tokenizers 0.15.0