swinv2-tiny-patch4-window8-256-DMAE-U2

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: 1.0453
• Accuracy: 0.6522

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: 3.5e-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: 0.05
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.86	3	1.3843	0.4783
No log	2.0	7	1.3617	0.4565
1.3721	2.86	10	1.3211	0.4565
1.3721	4.0	14	1.2598	0.4565
1.3721	4.86	17	1.2251	0.4565
1.25	6.0	21	1.2094	0.4565
1.25	6.86	24	1.2127	0.4565
1.25	8.0	28	1.2118	0.4565
1.1902	8.86	31	1.2031	0.4565
1.1902	10.0	35	1.1936	0.4565
1.1902	10.86	38	1.1829	0.4565
1.1472	12.0	42	1.1569	0.4565
1.1472	12.86	45	1.1432	0.4565
1.1472	14.0	49	1.1357	0.4783
1.1495	14.86	52	1.1178	0.5
1.1495	16.0	56	1.0903	0.5217
1.1495	16.86	59	1.0714	0.6304
1.0824	18.0	63	1.0453	0.6522
1.0824	18.86	66	1.0150	0.6522
1.0535	20.0	70	0.9925	0.6522
1.0535	20.86	73	0.9778	0.6304
1.0535	22.0	77	0.9570	0.6522
0.994	22.86	80	0.9441	0.6304
0.994	24.0	84	0.9246	0.6304
0.994	24.86	87	0.9095	0.6304
0.9554	26.0	91	0.8937	0.6304
0.9554	26.86	94	0.8925	0.6522
0.9554	28.0	98	0.8886	0.6522
0.953	28.86	101	0.8804	0.6522
0.953	30.0	105	0.8744	0.6522
0.953	30.86	108	0.8754	0.6522
0.9092	32.0	112	0.8733	0.6522
0.9092	32.86	115	0.8725	0.6522
0.9092	34.0	119	0.8722	0.6522
0.9247	34.29	120	0.8720	0.6522

Framework versions

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