swinv2-tiny-patch4-window8-256-DMAE-3e-5

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: 1.0126
• Accuracy: 0.6739

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: 3e-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	0.86	3	1.6337	0.1087
No log	2.0	7	1.6223	0.1087
1.6298	2.86	10	1.5972	0.1087
1.6298	4.0	14	1.5041	0.1087
1.6298	4.86	17	1.4149	0.1087
1.4858	6.0	21	1.3037	0.4565
1.4858	6.86	24	1.2491	0.4565
1.4858	8.0	28	1.2219	0.4565
1.2337	8.86	31	1.2146	0.4565
1.2337	10.0	35	1.2145	0.4565
1.2337	10.86	38	1.2116	0.4565
1.1726	12.0	42	1.2065	0.4565
1.1726	12.86	45	1.2011	0.4565
1.1726	14.0	49	1.1941	0.4565
1.1953	14.86	52	1.1878	0.4565
1.1953	16.0	56	1.1771	0.4565
1.1953	16.86	59	1.1677	0.4565
1.1602	18.0	63	1.1564	0.4565
1.1602	18.86	66	1.1502	0.4565
1.1415	20.0	70	1.1227	0.4565
1.1415	20.86	73	1.1017	0.4783
1.1415	22.0	77	1.0872	0.5870
1.0949	22.86	80	1.0795	0.5870
1.0949	24.0	84	1.0582	0.6087
1.0949	24.86	87	1.0402	0.6087
1.0559	26.0	91	1.0186	0.6522
1.0559	26.86	94	1.0126	0.6739
1.0559	28.0	98	1.0033	0.6739
1.0423	28.86	101	0.9849	0.6739
1.0423	30.0	105	0.9670	0.6739
1.0423	30.86	108	0.9631	0.6739
1.0002	32.0	112	0.9599	0.6522
1.0002	32.86	115	0.9587	0.6522
1.0002	34.0	119	0.9573	0.6522
1.0077	34.29	120	0.9571	0.6522

Framework versions

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