swinv2-tiny-patch4-window8-256-dmae-humeda-DAV52

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: 0.7948
• Accuracy: 0.7841

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• 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_with_restarts
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	6	1.6162	0.2045
1.3978	2.0	12	1.6140	0.1818
1.3978	3.0	18	1.4029	0.4318
1.0539	4.0	24	1.2503	0.5455
1.0539	5.0	30	1.0014	0.625
0.7171	6.0	36	0.9539	0.6364
0.7171	7.0	42	0.9958	0.6136
0.5557	8.0	48	0.8233	0.7386
0.5557	9.0	54	0.8813	0.6136
0.4942	10.0	60	0.8385	0.7159
0.4942	11.0	66	0.7914	0.7614
0.3957	12.0	72	0.7742	0.7273
0.3957	13.0	78	0.8122	0.7045
0.3664	14.0	84	0.7981	0.75
0.3664	15.0	90	0.7852	0.7159
0.3042	16.0	96	0.8829	0.7159
0.3042	17.0	102	0.7630	0.7386
0.2673	18.0	108	0.7936	0.75
0.2673	19.0	114	0.7491	0.7727
0.2308	20.0	120	0.7948	0.7841
0.2308	21.0	126	0.7798	0.7841
0.2113	22.0	132	0.7635	0.7614
0.2113	23.0	138	0.8521	0.7159
0.1852	24.0	144	0.8660	0.7386
0.1852	25.0	150	0.7984	0.7386
0.1765	26.0	156	0.7750	0.7614
0.1765	27.0	162	0.7935	0.7386
0.1969	28.0	168	0.7956	0.75
0.1969	29.0	174	0.7902	0.7727
0.1502	30.0	180	0.7868	0.7614
0.1502	31.0	186	0.7842	0.7614
0.1621	32.0	192	0.7836	0.75
0.1621	33.0	198	0.7837	0.75
0.1621	33.3810	200	0.7838	0.75

Framework versions

• Transformers 4.48.2
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0