swinv2-tiny-patch4-window8-256-dmae-humeda-DAV50

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: 0.8354
• Accuracy: 0.7273

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: 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	1.5451	0.3864
No log	2.0	10	1.5220	0.3864
1.4177	3.0	15	1.4938	0.4205
1.4177	4.0	20	1.4111	0.4432
1.2671	5.0	25	1.2941	0.4545
1.2671	6.0	30	1.2036	0.4545
1.2671	7.0	35	1.0816	0.5114
0.9869	8.0	40	1.0452	0.5795
0.9869	9.0	45	0.9876	0.625
0.8456	10.0	50	0.9791	0.5909
0.8456	11.0	55	0.9662	0.6023
0.7126	12.0	60	0.9302	0.6364
0.7126	13.0	65	0.9379	0.625
0.7126	14.0	70	0.9036	0.6705
0.6561	15.0	75	0.8846	0.6591
0.6561	16.0	80	0.8689	0.6591
0.6367	17.0	85	0.8543	0.6591
0.6367	18.0	90	0.8342	0.6932
0.6367	19.0	95	0.8185	0.6705
0.5463	20.0	100	0.8290	0.7159
0.5463	21.0	105	0.8354	0.7273
0.5504	22.0	110	0.8160	0.7159
0.5504	23.0	115	0.8073	0.7159
0.507	24.0	120	0.8071	0.7045
0.507	25.0	125	0.8071	0.6932
0.507	26.0	130	0.8047	0.7045
0.5226	27.0	135	0.8000	0.7045
0.5226	28.0	140	0.7987	0.7159
0.5144	29.0	145	0.8000	0.7159
0.5144	30.0	150	0.8002	0.7159
0.5144	31.0	155	0.8008	0.7159
0.4862	32.0	160	0.8008	0.7159

Framework versions

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