vit-base-patch16-224-RU4-40

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: 0.6467
• Accuracy: 0.8333

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: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.3822	0.99	19	1.3130	0.4833
1.2724	1.97	38	1.0987	0.6
0.9711	2.96	57	0.8624	0.6667
0.6349	4.0	77	0.7397	0.7333
0.4068	4.99	96	0.6979	0.75
0.2877	5.97	115	0.6270	0.7833
0.2217	6.96	134	0.6467	0.8333
0.195	8.0	154	0.6858	0.7833
0.1392	8.99	173	0.6505	0.8167
0.1534	9.97	192	0.6320	0.8167
0.1136	10.96	211	0.8346	0.7833
0.1025	12.0	231	0.6810	0.8
0.0894	12.99	250	0.8258	0.7667
0.1308	13.97	269	0.9456	0.75
0.0836	14.96	288	0.9084	0.8
0.0813	16.0	308	0.8688	0.8167
0.1017	16.99	327	0.8609	0.8
0.076	17.97	346	0.9015	0.8
0.0726	18.96	365	0.9918	0.7833
0.0549	20.0	385	0.9064	0.8
0.0676	20.99	404	0.8819	0.75
0.0717	21.97	423	0.8607	0.8167
0.0547	22.96	442	0.8859	0.8
0.0466	24.0	462	0.9328	0.8167
0.0715	24.99	481	1.0178	0.7667
0.0446	25.97	500	1.0094	0.7667
0.0468	26.96	519	0.9175	0.8167
0.0458	28.0	539	0.8580	0.8
0.0392	28.99	558	1.0589	0.7833
0.0469	29.97	577	1.0905	0.8
0.0425	30.96	596	1.0078	0.7833
0.0464	32.0	616	1.0206	0.7833
0.0336	32.99	635	0.9653	0.8167
0.0302	33.97	654	0.9574	0.8
0.0353	34.96	673	0.9621	0.8167
0.0344	36.0	693	0.9792	0.8167
0.0195	36.99	712	0.9459	0.8167
0.031	37.97	731	0.9488	0.8167
0.0224	38.96	750	0.9440	0.8167
0.0309	39.48	760	0.9448	0.8167

Framework versions

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