Using this script, we can implement several scenarios in audio classification, such as speaker identification, language recognition, emotion recognition, sentiment analysis and more, using Wav2vec2 and Whisper models.

For fine-tuning the Whisper model for audio classification: Whisper_Emotion.py

For fine-tuning the Wav2Vec2 for audio classification: wav2vec_Emotion.py

The manifest for feeding wav data must be like train_voice_emotion.csv file.

in addition we have wav2vec_Emotion_specaugm.py for utilizing SpecAugment as augmentation technique, wav2vec_embeding.py for extract and save feature embedings and wav2vec_emb_score.py for extracting scores for each wav file.

Please use slurm_run.sh to run the scripts with Slurm.