vit-base-patch16-224-RU3-40

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: 0.5667
• 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
1.3821	0.99	19	1.3119	0.4833
1.2698	1.97	38	1.0852	0.6167
0.9819	2.96	57	0.8757	0.7
0.6671	4.0	77	0.7689	0.7333
0.4248	4.99	96	0.7294	0.7167
0.3005	5.97	115	0.6518	0.7833
0.2035	6.96	134	0.5667	0.8333
0.2195	8.0	154	0.6646	0.8333
0.1654	8.99	173	0.6294	0.8167
0.1581	9.97	192	0.7211	0.7833
0.1338	10.96	211	0.8129	0.7833
0.1188	12.0	231	0.7925	0.8167
0.1179	12.99	250	0.9588	0.7667
0.1017	13.97	269	1.0875	0.7167
0.0845	14.96	288	0.9355	0.7
0.1109	16.0	308	0.9387	0.8167
0.0711	16.99	327	1.1214	0.7333
0.0884	17.97	346	0.9688	0.7667
0.0668	18.96	365	1.0306	0.8
0.0716	20.0	385	1.2653	0.7167
0.0643	20.99	404	0.9894	0.7833
0.0517	21.97	423	1.0439	0.7667
0.0597	22.96	442	1.1470	0.7667
0.0533	24.0	462	1.0848	0.7833
0.0529	24.99	481	1.1481	0.75
0.0524	25.97	500	1.1322	0.7333
0.0525	26.96	519	1.1868	0.7333
0.0517	28.0	539	1.1561	0.7167
0.0309	28.99	558	1.0562	0.7833
0.0403	29.97	577	1.2901	0.7333
0.0392	30.96	596	1.1295	0.7667
0.0404	32.0	616	1.1198	0.7667
0.0381	32.99	635	1.2986	0.7167
0.0262	33.97	654	1.1655	0.75
0.0354	34.96	673	1.1223	0.7833
0.0224	36.0	693	1.1679	0.7833
0.0244	36.99	712	1.0999	0.8167
0.0368	37.97	731	1.1213	0.7833
0.0199	38.96	750	1.1003	0.8
0.028	39.48	760	1.0989	0.8

Framework versions

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