swinv2-tiny-patch4-window8-256-RH-6e-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: 0.6666
• Accuracy: 0.6355

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: 6e-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	1.0	8	4.6188	0.4112
4.5343	2.0	16	4.4096	0.4112
4.5396	3.0	24	3.6784	0.4112
3.621	4.0	32	2.4800	0.4112
2.1758	5.0	40	1.2118	0.4112
2.1758	6.0	48	0.6790	0.5888
0.8871	7.0	56	0.7903	0.5888
0.7484	8.0	64	0.7640	0.5888
0.7414	9.0	72	0.6789	0.5888
0.6868	10.0	80	0.6770	0.5888
0.6868	11.0	88	0.6775	0.5888
0.6771	12.0	96	0.6993	0.5888
0.7082	13.0	104	0.6765	0.5888
0.6993	14.0	112	0.6746	0.5888
0.6798	15.0	120	0.6759	0.5888
0.6798	16.0	128	0.6734	0.5888
0.6734	17.0	136	0.6739	0.5888
0.6832	18.0	144	0.7039	0.5888
0.6825	19.0	152	0.6767	0.5888
0.6663	20.0	160	0.6707	0.5888
0.6663	21.0	168	0.6798	0.5888
0.6646	22.0	176	0.6723	0.5794
0.6764	23.0	184	0.6889	0.5888
0.6808	24.0	192	0.6994	0.5888
0.6766	25.0	200	0.6691	0.5888
0.6766	26.0	208	0.6837	0.5888
0.6698	27.0	216	0.6738	0.5701
0.6549	28.0	224	0.6695	0.5794
0.6442	29.0	232	0.7157	0.5794
0.649	30.0	240	0.6726	0.6075
0.649	31.0	248	0.6839	0.5794
0.6388	32.0	256	0.6797	0.5888
0.6416	33.0	264	0.6714	0.5981
0.6398	34.0	272	0.6730	0.6075
0.6522	35.0	280	0.6953	0.5794
0.6522	36.0	288	0.6609	0.5701
0.6376	37.0	296	0.6619	0.5794
0.6441	38.0	304	0.6654	0.6262
0.6149	39.0	312	0.6666	0.6355
0.623	40.0	320	0.6679	0.6355

Framework versions

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