swinv2-tiny-patch4-window8-256-dmae-humeda-DAV51

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: 0.6495
• Accuracy: 0.8182

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	5	1.5572	0.3409
No log	2.0	10	1.3890	0.4545
1.4186	3.0	15	1.2638	0.5795
1.4186	4.0	20	1.0291	0.6023
1.0199	5.0	25	1.0125	0.5909
1.0199	6.0	30	0.8328	0.6477
1.0199	7.0	35	0.8662	0.625
0.7093	8.0	40	0.7048	0.7045
0.7093	9.0	45	0.8032	0.6818
0.576	10.0	50	0.6944	0.7273
0.576	11.0	55	0.7730	0.6932
0.4817	12.0	60	0.6605	0.7386
0.4817	13.0	65	0.7316	0.75
0.4817	14.0	70	0.6380	0.7727
0.413	15.0	75	0.6573	0.7727
0.413	16.0	80	0.6570	0.75
0.3959	17.0	85	0.6173	0.7955
0.3959	18.0	90	0.6293	0.7841
0.3959	19.0	95	0.6491	0.7727
0.3043	20.0	100	0.6382	0.7955
0.3043	21.0	105	0.6272	0.7955
0.295	22.0	110	0.6423	0.8068
0.295	23.0	115	0.6413	0.8068
0.2365	24.0	120	0.6388	0.7841
0.2365	25.0	125	0.6457	0.7841
0.2365	26.0	130	0.6513	0.7955
0.2507	27.0	135	0.6495	0.8182
0.2507	28.0	140	0.6463	0.8182
0.2385	29.0	145	0.6468	0.8068
0.2385	30.0	150	0.6480	0.8068
0.2385	31.0	155	0.6484	0.8068
0.2432	32.0	160	0.6486	0.8068

Framework versions

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