SWv2-DMAE-H-4-rp-clean-fix-U-40-Cross-2

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: 0.4464
• Accuracy: 0.8929

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
1.6089	0.98	12	1.6061	0.2024
1.6027	1.96	24	1.5770	0.2024
1.563	2.94	36	1.5561	0.2024
1.5137	4.0	49	1.4363	0.2024
1.3706	4.98	61	1.2349	0.6548
1.2475	5.96	73	0.9802	0.7857
1.0611	6.94	85	0.7390	0.7857
0.9738	8.0	98	0.5656	0.8333
0.8107	8.98	110	0.4464	0.8929
0.7725	9.96	122	0.4132	0.8690
0.6953	10.94	134	0.4220	0.8214
0.6906	12.0	147	0.4704	0.8214
0.6945	12.98	159	0.4340	0.8690
0.6033	13.96	171	0.3935	0.8452
0.5701	14.94	183	0.4422	0.8452
0.5709	16.0	196	0.6004	0.7262
0.5525	16.98	208	0.4029	0.8452
0.5094	17.96	220	0.3620	0.8690
0.485	18.94	232	0.3388	0.8929
0.4659	20.0	245	0.3637	0.8452
0.4763	20.98	257	0.4209	0.8333
0.3919	21.96	269	0.3549	0.8690
0.4033	22.94	281	0.3585	0.8571
0.382	24.0	294	0.3584	0.8452
0.3684	24.98	306	0.3661	0.8690
0.3513	25.96	318	0.3481	0.8929
0.3909	26.94	330	0.3629	0.8929
0.3235	28.0	343	0.3703	0.8929
0.3184	28.98	355	0.3469	0.8810
0.3401	29.96	367	0.3633	0.8929
0.413	30.94	379	0.3611	0.8690
0.353	32.0	392	0.3607	0.8810
0.2911	32.98	404	0.3436	0.8929
0.3263	33.96	416	0.3357	0.8810
0.307	34.94	428	0.3385	0.8929
0.322	36.0	441	0.3441	0.8810
0.2888	36.98	453	0.3407	0.8690
0.2862	37.96	465	0.3390	0.8690
0.2937	38.94	477	0.3386	0.8690
0.2893	39.18	480	0.3386	0.8690

Framework versions

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