vit-Facial-Expression-Recognition

This model is a fine-tuned version of motheecreator/vit-Facial-Expression-Recognition on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3658
• Accuracy: 0.8753
• F1: 0.8737
• Precision: 0.8749
• Recall: 0.8753

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 256
• optimizer: Use 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_steps: 1000
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
4.5618	0.2164	100	0.3710	0.8762	0.8746	0.8752	0.8762
4.6091	0.4328	200	0.3677	0.8761	0.8747	0.8762	0.8761
4.5423	0.6492	300	0.3695	0.8748	0.8730	0.8745	0.8748
4.6307	0.8656	400	0.3745	0.8711	0.8692	0.8730	0.8711
4.3953	1.0801	500	0.3745	0.8727	0.8711	0.8724	0.8727
4.341	1.2965	600	0.3803	0.8688	0.8674	0.8688	0.8688
4.5471	1.5128	700	0.3841	0.8713	0.8699	0.8710	0.8713
4.522	1.7292	800	0.3836	0.8679	0.8662	0.8678	0.8679
4.5596	1.9456	900	0.3885	0.8672	0.8649	0.8678	0.8672
4.1491	2.1601	1000	0.3849	0.8691	0.8677	0.8689	0.8691
4.1037	2.3765	1100	0.3906	0.8667	0.8647	0.8669	0.8667
4.0033	2.5929	1200	0.3784	0.8704	0.8687	0.8699	0.8704
3.9759	2.8093	1300	0.3677	0.8752	0.8737	0.8747	0.8752

Framework versions

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