mixed_model_finetuned_cremad

This model is a fine-tuned version of wav2vec2 on audio stream part and pretrained resnet3d_101 on video stream part , It was trained from scratch on CremaD dataset. This dataset provides 7442 samples of recordings from actors performing on 6 different emotions in English, which are:

emotions = ['angry', 'disgust', 'fearful', 'happy', 'neutral', 'sad']

It achieves the following results on the evaluation set:

• Loss: 0.3098
• Accuracy: 0.8972
• F1: 0.8960
• Recall: 0.8972
• Precision: 0.8974

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: 0.0001
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 743
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Recall	Precision
0.7914	1.0	186	1.0595	0.7171	0.7074	0.7171	0.7536
0.5971	2.0	372	0.4401	0.8414	0.8375	0.8414	0.8443
0.2891	3.0	558	0.3863	0.8548	0.8539	0.8548	0.8622
0.1833	3.9946	743	0.3098	0.8972	0.8960	0.8972	0.8974

Framework versions

• Transformers 4.42.3
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1