SWv2-DMAE-H-5-p-clean-fix-U-40-Cross-3

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: 0.4952
• Accuracy: 0.8391

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
1.609	0.96	13	1.6081	0.1954
1.6048	2.0	27	1.5943	0.1954
1.5299	2.96	40	1.5695	0.1954
1.4527	4.0	54	1.4408	0.2184
1.3442	4.96	67	1.1869	0.5632
1.114	6.0	81	0.9027	0.6897
0.9651	6.96	94	0.7102	0.7586
0.8893	8.0	108	0.8029	0.6322
0.7704	8.96	121	0.5880	0.7816
0.6737	10.0	135	0.5514	0.8161
0.6713	10.96	148	0.4952	0.8391
0.6927	12.0	162	0.5375	0.8046
0.6031	12.96	175	0.5099	0.7931
0.542	14.0	189	0.5453	0.8046
0.511	14.96	202	0.5673	0.7701
0.4901	16.0	216	0.6610	0.7931
0.4824	16.96	229	0.5610	0.8046
0.4685	18.0	243	0.5992	0.7931
0.4442	18.96	256	0.5958	0.8161
0.4676	20.0	270	0.5621	0.8391
0.4231	20.96	283	0.5877	0.8161
0.3795	22.0	297	0.6075	0.8046
0.3645	22.96	310	0.6449	0.8046
0.366	24.0	324	0.6480	0.8161
0.344	24.96	337	0.6409	0.8276
0.2833	26.0	351	0.6246	0.8161
0.326	26.96	364	0.6491	0.8276
0.3416	28.0	378	0.6778	0.8161
0.2942	28.96	391	0.6374	0.8276
0.2767	30.0	405	0.6426	0.8276
0.2569	30.96	418	0.6631	0.8276
0.2889	32.0	432	0.6818	0.8161
0.2701	32.96	445	0.6889	0.8046
0.2553	34.0	459	0.6986	0.8161
0.2401	34.96	472	0.6985	0.8046
0.2536	36.0	486	0.7015	0.8046
0.2738	36.96	499	0.6940	0.8046
0.2304	38.0	513	0.6951	0.8046
0.3081	38.52	520	0.6951	0.8046

Framework versions

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