swinv2-tiny-patch4-window8-256-RH

This model is a fine-tuned version of microsoft/swinv2-tiny-patch4-window8-256 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.6396
• Accuracy: 0.6822

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: 4e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	8	4.6265	0.4112
4.5369	2.0	16	4.5295	0.4112
4.6305	3.0	24	4.1439	0.4112
4.0918	4.0	32	3.3693	0.4112
3.1767	5.0	40	2.4325	0.4112
3.1767	6.0	48	1.5422	0.4112
2.0113	7.0	56	0.8834	0.4112
1.0593	8.0	64	0.6790	0.5888
0.696	9.0	72	0.7044	0.5888
0.6893	10.0	80	0.6778	0.5888
0.6893	11.0	88	0.6866	0.5888
0.6961	12.0	96	0.6934	0.5888
0.7329	13.0	104	0.6915	0.5888
0.6948	14.0	112	0.6762	0.5888
0.6771	15.0	120	0.6795	0.5888
0.6771	16.0	128	0.6801	0.5888
0.6763	17.0	136	0.6820	0.5888
0.6822	18.0	144	0.6800	0.5888
0.6723	19.0	152	0.6741	0.5888
0.6757	20.0	160	0.6815	0.5888
0.6757	21.0	168	0.6729	0.5888
0.6711	22.0	176	0.6812	0.5888
0.6784	23.0	184	0.6781	0.5794
0.6665	24.0	192	0.6698	0.5794
0.6723	25.0	200	0.6647	0.5981
0.6723	26.0	208	0.6762	0.5794
0.6675	27.0	216	0.6597	0.5701
0.6628	28.0	224	0.6563	0.6355
0.6478	29.0	232	0.6791	0.5794
0.6642	30.0	240	0.6574	0.5888
0.6642	31.0	248	0.6556	0.5607
0.654	32.0	256	0.6523	0.5888
0.6602	33.0	264	0.6464	0.6262
0.6535	34.0	272	0.6450	0.6168
0.6506	35.0	280	0.6550	0.5794
0.6506	36.0	288	0.6438	0.6075
0.6533	37.0	296	0.6396	0.6822
0.6443	38.0	304	0.6383	0.6636
0.6263	39.0	312	0.6378	0.6449
0.6283	40.0	320	0.6379	0.6449

Framework versions

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