mirror of
https://github.com/modelscope/FunASR
synced 2025-09-15 14:48:36 +08:00
14 lines
511 B
Python
14 lines
511 B
Python
import torch
|
|
from pathlib import Path
|
|
from funasr_torch.paraformer_bin import ContextualParaformer
|
|
|
|
model_dir = "iic/speech_paraformer-large-contextual_asr_nat-zh-cn-16k-common-vocab8404"
|
|
device_id = 0 if torch.cuda.is_available() else -1
|
|
model = ContextualParaformer(model_dir, batch_size=1, device_id=device_id) # gpu
|
|
|
|
wav_path = "{}/.cache/modelscope/hub/{}/example/asr_example.wav".format(Path.home(), model_dir)
|
|
hotwords = "你的热词 魔搭 达摩苑"
|
|
|
|
result = model(wav_path, hotwords)
|
|
print(result)
|