gizachewteshome/NLPIR-Amharic-SER

This is NLPIR course project on Fine-tuning Pre-trained models for Amharic Speech Emotion Recogntion. The model is a fine-tuned version of Wev2Vec2.0 and HuBERT base & large vesrion on the ASED dataset. To use fine-tuned Amharic SER model: # Wev2Vec2.0 from transformers import AutoProcessor, Wav2Vec2ForSpeechClassification processor = AutoProcessor.from_pretrained("Gizachew/wev2vec-base960-agu-amharic") model = Wav2Vec2ForSpeechClassification.from_pretrained("Gizachew/wev2vec-base960-agu-amharic") Wev2Vec2.0-base: Gizachew/wev2vec-base960-ag… Wev2Vec2.0-large: Gizachew/wev2vec-large960-a… # HuBERT from transformers import AutoProcessor, HubertForSpeechClassification processor = AutoProcessor.from_pretrained("Gizachew/hubert-agum960-amharic") model = HubertForSpeechClassification.from_pretrained("Gizachew/hubert-agum960-amharic") HuBERT-base: Gizachew/hubert-base960-agu… HuBERT-large: Gizachew/hubert-agum960-amh… Demo: huggingface.co The orginal dataset used for this experiment is in this link: github.com