Views
No views yet
1import transformers
2import numpy as np
3import librosa
4
5pipe = transformers.pipeline(model='AtAndDev/UVOX-50k-Llama-3.2-1B-Instruct', trust_remote_code=True, device="cuda")
6
7path = "voice_input.mp3"
8audio, sr = librosa.load(path, sr=16000)
9
10turns = []
11pipe({'audio': audio, 'turns': turns, 'sampling_rate': sr}, max_new_tokens=100)