ReVoiceAI-swinv2-base-512-face-rehab-finetuned

This model is a fine-tuned version of microsoft/swinv2-base-patch4-window12-192-22k on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.9237
• Accuracy: 0.8429
• F1: 0.8356

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: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 40
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.5461	1.0	57	1.1784	0.7484	0.7068
0.5285	2.0	114	1.3700	0.6667	0.6434
0.5202	3.0	171	0.9857	0.7917	0.7692
0.5172	4.0	228	1.4417	0.6458	0.6336
0.5174	5.0	285	1.6251	0.6074	0.5680
0.5127	6.0	342	1.5429	0.6170	0.6015
0.5187	7.0	399	1.3860	0.6939	0.6835
0.5171	8.0	456	0.9721	0.8141	0.8075
0.5105	9.0	513	1.4036	0.6522	0.6183
0.4872	10.0	570	1.1586	0.7885	0.7792
0.5148	11.0	627	1.0491	0.7997	0.7794
0.5211	12.0	684	0.9976	0.8285	0.8222
0.515	13.0	741	1.5812	0.6170	0.6101
0.5076	14.0	798	0.9237	0.8429	0.8356

Framework versions

• Transformers 4.53.0
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.2