#!/usr/bin/env python3
|
# -*- encoding: utf-8 -*-
|
# Copyright FunASR (https://github.com/FunAudioLLM/SenseVoice). All Rights Reserved.
|
# MIT License (https://opensource.org/licenses/MIT)
|
|
|
import torch
|
import os.path
|
import librosa
|
import numpy as np
|
from pathlib import Path
|
from typing import List, Union, Tuple
|
|
from .utils.utils import (
|
CharTokenizer,
|
get_logger,
|
read_yaml,
|
)
|
from .utils.frontend import WavFrontend
|
from .utils.sentencepiece_tokenizer import SentencepiecesTokenizer
|
|
logging = get_logger()
|
|
|
class SenseVoiceSmall:
|
"""
|
Author: Speech Lab of DAMO Academy, Alibaba Group
|
Paraformer: Fast and Accurate Parallel Transformer for Non-autoregressive End-to-End Speech Recognition
|
https://arxiv.org/abs/2206.08317
|
"""
|
|
def __init__(
|
self,
|
model_dir: Union[str, Path] = None,
|
batch_size: int = 1,
|
plot_timestamp_to: str = "",
|
quantize: bool = False,
|
intra_op_num_threads: int = 4,
|
cache_dir: str = None,
|
**kwargs,
|
):
|
self.device = kwargs.get("device", "cpu")
|
if not Path(model_dir).exists():
|
try:
|
from modelscope.hub.snapshot_download import snapshot_download
|
except:
|
raise "You are exporting model from modelscope, please install modelscope and try it again. To install modelscope, you could:\n" "\npip3 install -U modelscope\n" "For the users in China, you could install with the command:\n" "\npip3 install -U modelscope -i https://mirror.sjtu.edu.cn/pypi/web/simple"
|
try:
|
model_dir = snapshot_download(model_dir, cache_dir=cache_dir)
|
except:
|
raise "model_dir must be model_name in modelscope or local path downloaded from modelscope, but is {}".format(
|
model_dir
|
)
|
|
model_file = os.path.join(model_dir, "model.torchscript")
|
if quantize:
|
model_file = os.path.join(model_dir, "model_quant.torchscript")
|
if not os.path.exists(model_file):
|
print(".torchscripts does not exist, begin to export torchscript")
|
try:
|
from funasr import AutoModel
|
except:
|
raise "You are exporting onnx, please install funasr and try it again. To install funasr, you could:\n" "\npip3 install -U funasr\n" "For the users in China, you could install with the command:\n" "\npip3 install -U funasr -i https://mirror.sjtu.edu.cn/pypi/web/simple"
|
|
model = AutoModel(model=model_dir)
|
model_dir = model.export(type="torchscript", quantize=quantize, **kwargs)
|
|
config_file = os.path.join(model_dir, "config.yaml")
|
cmvn_file = os.path.join(model_dir, "am.mvn")
|
config = read_yaml(config_file)
|
|
self.tokenizer = SentencepiecesTokenizer(
|
bpemodel=os.path.join(model_dir, "chn_jpn_yue_eng_ko_spectok.bpe.model")
|
)
|
config["frontend_conf"]["cmvn_file"] = cmvn_file
|
self.frontend = WavFrontend(**config["frontend_conf"])
|
self.ort_infer = torch.jit.load(model_file)
|
self.batch_size = batch_size
|
self.blank_id = 0
|
self.lid_dict = {"auto": 0, "zh": 3, "en": 4, "yue": 7, "ja": 11, "ko": 12, "nospeech": 13}
|
self.lid_int_dict = {24884: 3, 24885: 4, 24888: 7, 24892: 11, 24896: 12, 24992: 13}
|
self.textnorm_dict = {"withitn": 14, "woitn": 15}
|
self.textnorm_int_dict = {25016: 14, 25017: 15}
|
|
def __call__(self, wav_content: Union[str, np.ndarray, List[str]], **kwargs) -> List:
|
|
language = self.lid_dict[kwargs.get("language", "auto")]
|
use_itn = kwargs.get("use_itn", False)
|
textnorm = kwargs.get("text_norm", None)
|
if textnorm is None:
|
textnorm = "withitn" if use_itn else "woitn"
|
textnorm = self.textnorm_dict[textnorm]
|
|
waveform_list = self.load_data(wav_content, self.frontend.opts.frame_opts.samp_freq)
|
waveform_nums = len(waveform_list)
|
asr_res = []
|
for beg_idx in range(0, waveform_nums, self.batch_size):
|
end_idx = min(waveform_nums, beg_idx + self.batch_size)
|
feats, feats_len = self.extract_feat(waveform_list[beg_idx:end_idx])
|
ctc_logits, encoder_out_lens = self.ort_infer(
|
torch.Tensor(feats).to(self.device),
|
torch.Tensor(feats_len).to(self.device),
|
torch.tensor([language]).to(self.device),
|
torch.tensor([textnorm]).to(self.device),
|
)
|
# support batch_size=1 only currently
|
x = ctc_logits[0, : encoder_out_lens[0].item(), :]
|
yseq = x.argmax(dim=-1)
|
yseq = torch.unique_consecutive(yseq, dim=-1)
|
|
mask = yseq != self.blank_id
|
token_int = yseq[mask].tolist()
|
|
if tokenizer is not None:
|
asr_res.append(self.tokenizer.decode(token_int))
|
else:
|
asr_res.append(token_int)
|
return asr_res
|
|
def load_data(self, wav_content: Union[str, np.ndarray, List[str]], fs: int = None) -> List:
|
def load_wav(path: str) -> np.ndarray:
|
waveform, _ = librosa.load(path, sr=fs)
|
return waveform
|
|
if isinstance(wav_content, np.ndarray):
|
return [wav_content]
|
|
if isinstance(wav_content, str):
|
return [load_wav(wav_content)]
|
|
if isinstance(wav_content, list):
|
return [load_wav(path) for path in wav_content]
|
|
raise TypeError(f"The type of {wav_content} is not in [str, np.ndarray, list]")
|
|
def extract_feat(self, waveform_list: List[np.ndarray]) -> Tuple[np.ndarray, np.ndarray]:
|
feats, feats_len = [], []
|
for waveform in waveform_list:
|
speech, _ = self.frontend.fbank(waveform)
|
feat, feat_len = self.frontend.lfr_cmvn(speech)
|
feats.append(feat)
|
feats_len.append(feat_len)
|
|
feats = self.pad_feats(feats, np.max(feats_len))
|
feats_len = np.array(feats_len).astype(np.int32)
|
return feats, feats_len
|
|
@staticmethod
|
def pad_feats(feats: List[np.ndarray], max_feat_len: int) -> np.ndarray:
|
def pad_feat(feat: np.ndarray, cur_len: int) -> np.ndarray:
|
pad_width = ((0, max_feat_len - cur_len), (0, 0))
|
return np.pad(feat, pad_width, "constant", constant_values=0)
|
|
feat_res = [pad_feat(feat, feat.shape[0]) for feat in feats]
|
feats = np.array(feat_res).astype(np.float32)
|
return feats
|
