swinv2-tiny-patch4-window8-256-DMAE-15e

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: 7.9238
• Accuracy: 0.1087

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: 1.5e-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
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.86	3	7.9238	0.1087
No log	2.0	7	7.8746	0.1087
7.9273	2.86	10	7.8185	0.1087
7.9273	4.0	14	7.6996	0.1087
7.9273	4.86	17	7.5876	0.1087
7.6529	6.0	21	7.4189	0.1087
7.6529	6.86	24	7.2678	0.1087
7.6529	8.0	28	7.0489	0.1087
7.1057	8.86	31	6.8846	0.1087
7.1057	10.0	35	6.6868	0.1087
7.1057	10.86	38	6.5595	0.1087
6.8483	12.0	42	6.3826	0.1087
6.8483	12.86	45	6.2276	0.1087
6.8483	14.0	49	6.0366	0.1087
6.224	14.86	52	5.9044	0.1087
6.224	16.0	56	5.7383	0.1087
6.224	16.86	59	5.6266	0.1087
5.8234	18.0	63	5.4871	0.1087
5.8234	18.86	66	5.3891	0.1087
5.5423	20.0	70	5.2672	0.1087
5.5423	20.86	73	5.1809	0.1087
5.5423	22.0	77	5.0741	0.1087
5.2547	22.86	80	5.0007	0.1087
5.2547	24.0	84	4.9116	0.1087
5.2547	24.86	87	4.8505	0.1087
5.0166	26.0	91	4.7770	0.1087
5.0166	26.86	94	4.7281	0.1087
5.0166	28.0	98	4.6712	0.1087
4.7751	28.86	101	4.6348	0.1087
4.7751	30.0	105	4.5943	0.1087
4.7751	30.86	108	4.5701	0.1087
4.7321	32.0	112	4.5458	0.1087
4.7321	32.86	115	4.5336	0.1087
4.7321	34.0	119	4.5255	0.1087
4.6731	34.29	120	4.5249	0.1087

Framework versions

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