swinv2-tiny-patch4-window8-256-dmae-humeda-DAV35

This model is a fine-tuned version of microsoft/swinv2-tiny-patch4-window8-256 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.2578
• Accuracy: 0.7

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.85	1.0	36	1.4133	0.5333
1.9294	2.0	72	0.9294	0.6333
1.1818	3.0	108	0.7700	0.65
0.7534	4.0	144	0.7531	0.7167
0.4285	5.0	180	0.9580	0.7
0.08	6.0	216	1.1785	0.75
0.0891	7.0	252	1.4686	0.7333
0.0602	8.0	288	1.7816	0.7
0.0284	9.0	324	1.5790	0.7667
0.0513	10.0	360	1.8933	0.7
0.0335	11.0	396	2.1433	0.65
0.025	12.0	432	2.3483	0.6667
0.0246	13.0	468	2.6426	0.6667
0.0306	14.0	504	3.0153	0.65
0.016	15.0	540	3.1259	0.6833
0.006	16.0	576	2.7612	0.7167
0.0234	17.0	612	2.5334	0.7167
0.0025	18.0	648	2.1768	0.7667
0.0001	19.0	684	2.6585	0.7167
0.0007	20.0	720	2.3282	0.7167
0.0003	21.0	756	2.6975	0.7333
0.0003	22.0	792	2.6186	0.7
0.0006	23.0	828	2.9600	0.7167
0.0008	24.0	864	2.9623	0.7333
0.0002	25.0	900	2.8632	0.7167
0.0143	26.0	936	2.8460	0.7167
0.0	27.0	972	2.9372	0.7167
0.0002	28.0	1008	2.8056	0.75
0.0001	29.0	1044	3.0591	0.7167
0.0001	30.0	1080	3.3295	0.6833
0.0	31.0	1116	3.2851	0.6833
0.0001	32.0	1152	3.4065	0.7
0.0	33.0	1188	3.3669	0.7
0.0	34.0	1224	3.3185	0.7167
0.0006	35.0	1260	3.2563	0.7
0.0004	36.0	1296	3.2831	0.7
0.0001	37.0	1332	3.2594	0.7
0.0	38.0	1368	3.2576	0.7
0.0	38.9014	1400	3.2578	0.7

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0