SW2-RHS-DA

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: 0.4541
• Accuracy: 0.8598

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
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.0538	0.99	35	1.2866	0.4112
0.7464	2.0	71	0.6780	0.5888
0.7061	2.99	106	0.6851	0.5888
0.6951	4.0	142	0.6742	0.5888
0.6928	4.99	177	0.6917	0.4486
0.683	6.0	213	0.6531	0.5794
0.7013	6.99	248	0.7000	0.4486
0.6921	8.0	284	0.7519	0.5514
0.6166	8.99	319	0.5947	0.6822
0.6128	10.0	355	0.5434	0.7850
0.5737	10.99	390	0.5533	0.7570
0.5376	12.0	426	0.5347	0.7103
0.5056	12.99	461	0.4949	0.7664
0.5396	14.0	497	0.5151	0.7477
0.4826	14.99	532	0.5669	0.7196
0.4269	16.0	568	0.4796	0.7570
0.5004	16.99	603	0.4489	0.8037
0.4116	18.0	639	0.4362	0.8224
0.3776	18.99	674	0.5300	0.7570
0.3646	20.0	710	0.4175	0.8037
0.3683	20.99	745	0.4700	0.8224
0.3277	22.0	781	0.4707	0.8131
0.3534	22.99	816	0.5240	0.8131
0.3083	24.0	852	0.5012	0.8131
0.2829	24.99	887	0.4421	0.8318
0.2564	26.0	923	0.4548	0.8224
0.3136	26.99	958	0.4374	0.8318
0.2443	28.0	994	0.5277	0.8131
0.258	28.99	1029	0.4601	0.8224
0.2673	30.0	1065	0.4520	0.8318
0.2233	30.99	1100	0.4541	0.8598
0.2276	32.0	1136	0.4247	0.8505
0.2653	32.99	1171	0.4091	0.8505
0.2007	34.0	1207	0.4719	0.8505
0.2082	34.99	1242	0.4624	0.8411
0.1794	36.0	1278	0.4856	0.8318
0.1987	36.99	1313	0.4904	0.8224
0.2066	38.0	1349	0.4741	0.8505
0.1972	38.99	1384	0.4530	0.8505
0.2319	39.44	1400	0.4536	0.8505

Framework versions

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