|
from typing import Dict, List, Any |
|
from transformers import AutoProcessor, MusicgenForConditionalGeneration |
|
import torch |
|
import array |
|
import base64 |
|
import io |
|
import wave |
|
import numpy as np |
|
|
|
class EndpointHandler: |
|
def __init__(self, path=""): |
|
|
|
self.processor = AutoProcessor.from_pretrained(path) |
|
self.model = MusicgenForConditionalGeneration.from_pretrained(path, torch_dtype=torch.float16).to("cuda") |
|
|
|
def __call__(self, data: Dict[str, Any]) -> Dict[str, str]: |
|
""" |
|
Args: |
|
data (:dict:): |
|
The payload with the text prompt and generation parameters. |
|
""" |
|
|
|
inputs = data.pop("inputs", data) |
|
parameters = data.pop("parameters", None) |
|
|
|
|
|
inputs = self.processor( |
|
text=[inputs], |
|
padding=True, |
|
return_tensors="pt",).to("cuda") |
|
|
|
|
|
with torch.autocast("cuda"): |
|
audio_values = self.model.generate(**inputs, do_sample=True, guidance_scale=3, max_new_tokens=400) |
|
|
|
|
|
sampling_rate = self.model.config.audio_encoder.sampling_rate |
|
audio_samples = audio_values[0].cpu().numpy()[0].tolist() |
|
|
|
audio_samples = [int(min(max(sample * 32000, -32000), 32000)) for sample in audio_samples] |
|
|
|
|
|
audio_io = io.BytesIO() |
|
|
|
|
|
with wave.open(audio_io, 'wb') as wf: |
|
wf.setnchannels(1) |
|
wf.setsampwidth(2) |
|
wf.setframerate(sampling_rate) |
|
wf.writeframes(array.array('h', audio_samples).tobytes()) |
|
|
|
audio_base64 = base64.b64encode(audio_io.getvalue()).decode('utf-8') |
|
|
|
return [{'sampling_rate': sampling_rate, 'audio': audio_base64}] |
