swin-tiny-patch4-window7-224-dmae-va-da2-40

This model is a fine-tuned version of microsoft/swin-tiny-patch4-window7-224 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3718
• Accuracy: 0.8966

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: 0.1
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.94	4	1.3103	0.3621
No log	1.88	8	1.2844	0.3276
1.3144	2.82	12	1.2575	0.4138
1.3144	4.0	17	1.1181	0.5
1.1723	4.94	21	1.0054	0.5345
1.1723	5.88	25	0.8771	0.5690
1.1723	6.82	29	0.7941	0.7069
0.9243	8.0	34	0.7038	0.6724
0.9243	8.94	38	0.7341	0.7241
0.7394	9.88	42	0.6256	0.7414
0.7394	10.82	46	0.6059	0.7586
0.627	12.0	51	0.5662	0.7414
0.627	12.94	55	0.5384	0.7931
0.627	13.88	59	0.5631	0.7759
0.5026	14.82	63	0.4938	0.8103
0.5026	16.0	68	0.4768	0.8448
0.4284	16.94	72	0.4083	0.8621
0.4284	17.88	76	0.3942	0.8276
0.3922	18.82	80	0.3760	0.8621
0.3922	20.0	85	0.4109	0.8276
0.3922	20.94	89	0.3852	0.8621
0.3927	21.88	93	0.3661	0.8621
0.3927	22.82	97	0.3930	0.8793
0.3338	24.0	102	0.3586	0.8621
0.3338	24.94	106	0.3607	0.8621
0.2698	25.88	110	0.3718	0.8966
0.2698	26.82	114	0.3502	0.8793
0.2698	28.0	119	0.3614	0.8448
0.2707	28.94	123	0.3537	0.8621
0.2707	29.88	127	0.3545	0.8793
0.2622	30.82	131	0.3202	0.8793
0.2622	32.0	136	0.3125	0.8793
0.2499	32.94	140	0.3215	0.8621
0.2499	33.88	144	0.3194	0.8621
0.2499	34.82	148	0.3066	0.8793
0.2157	36.0	153	0.3024	0.8793
0.2157	36.94	157	0.3037	0.8793
0.2248	37.65	160	0.3035	0.8793

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0