vit-base-patch16-224-RU5-40

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: 0.6150
• Accuracy: 0.85

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: 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
1.3806	0.95	14	1.3385	0.4833
1.3323	1.97	29	1.1803	0.6
1.1086	2.98	44	0.9835	0.6333
0.927	4.0	59	0.8340	0.7167
0.6591	4.95	73	0.7843	0.7167
0.5201	5.97	88	0.7683	0.7167
0.3763	6.98	103	0.7880	0.6833
0.26	8.0	118	0.6876	0.7667
0.2219	8.95	132	0.7188	0.7833
0.2243	9.97	147	0.8730	0.7
0.178	10.98	162	0.6872	0.7833
0.1944	12.0	177	0.6150	0.85
0.1422	12.95	191	0.6832	0.7833
0.1117	13.97	206	0.7590	0.7833
0.117	14.98	221	0.8429	0.7667
0.1176	16.0	236	0.9741	0.7667
0.1081	16.95	250	0.9106	0.7833
0.0928	17.97	265	0.9179	0.7333
0.0848	18.98	280	0.9695	0.7667
0.1045	20.0	295	0.8805	0.8
0.1159	20.95	309	0.9458	0.7667
0.0748	21.97	324	0.8463	0.7667
0.0641	22.98	339	0.8815	0.8
0.0799	24.0	354	0.9426	0.75
0.0921	24.95	368	0.9212	0.75
0.0602	25.97	383	0.9828	0.75
0.059	26.98	398	0.8861	0.8
0.0669	28.0	413	0.9302	0.7333
0.0508	28.95	427	1.0306	0.7167
0.0585	29.97	442	0.9149	0.75
0.0619	30.98	457	0.8942	0.7833
0.0626	32.0	472	0.9069	0.7667
0.0575	32.95	486	0.8656	0.8
0.0483	33.97	501	0.8779	0.8167
0.0576	34.98	516	0.9078	0.7833
0.0633	36.0	531	0.8880	0.8
0.0511	36.95	545	0.8573	0.8
0.049	37.97	560	0.8564	0.8

Framework versions

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