Wav2Vec2-MultiTask

This is a fine-tuned Wav2Vec2.0 model for multi-task learning:

• Phoneme recognition
• Emotion classification
• Speaker identification

Usage

from transformers import AutoModel, AutoConfig, AutoProcessor

model = AutoModel.from_pretrained(
    "username/my-wav2vec2-multitask",
    trust_remote_code=True
)

config = AutoConfig.from_pretrained(
    "username/my-wav2vec2-multitask",
    trust_remote_code=True
)

processor = AutoProcessor.from_pretrained("facebook/wav2vec2-base")

inputs = processor("hello world", return_tensors="pt", sampling_rate=16000)

# phoneme recognition
logits = model(**inputs, task="phoneme")