AtAndDev/UVOX-50k-Llama-3.2-1B-Instruct

import transformers
import numpy as np
import librosa

pipe = transformers.pipeline(model='AtAndDev/UVOX-50k-Llama-3.2-1B-Instruct', trust_remote_code=True, device="cuda")

path = "voice_input.mp3"
audio, sr = librosa.load(path, sr=16000)

turns = []
pipe({'audio': audio, 'turns': turns, 'sampling_rate': sr}, max_new_tokens=100)