From a62cd7a3fdae0e48da16558daf89a4498512fdb9 Mon Sep 17 00:00:00 2001
From: shixian <shixian@U-09RYG5WD-2244.local>
Date: 星期五, 20 十二月 2024 10:50:51 +0800
Subject: [PATCH] update sensevoice onnx
---
runtime/python/onnxruntime/funasr_onnx/sensevoice_bin.py | 29 ++++++++++++++++++++++++-----
1 files changed, 24 insertions(+), 5 deletions(-)
diff --git a/runtime/python/onnxruntime/funasr_onnx/sensevoice_bin.py b/runtime/python/onnxruntime/funasr_onnx/sensevoice_bin.py
index 37c98a8..bf325c8 100644
--- a/runtime/python/onnxruntime/funasr_onnx/sensevoice_bin.py
+++ b/runtime/python/onnxruntime/funasr_onnx/sensevoice_bin.py
@@ -3,8 +3,6 @@
# 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
@@ -181,12 +179,12 @@
)
for b in range(feats.shape[0]):
# back to torch.Tensor
- if isinstance(ctc_logits, np.ndarray):
- ctc_logits = torch.from_numpy(ctc_logits).float()
+ # if isinstance(ctc_logits, np.ndarray):
+ # ctc_logits = torch.from_numpy(ctc_logits).float()
# support batch_size=1 only currently
x = ctc_logits[b, : encoder_out_lens[b].item(), :]
yseq = x.argmax(dim=-1)
- yseq = torch.unique_consecutive(yseq, dim=-1)
+ yseq = np.unique(yseq)
mask = yseq != self.blank_id
token_int = yseq[mask].tolist()
@@ -196,7 +194,24 @@
return asr_res
def load_data(self, wav_content: Union[str, np.ndarray, List[str]], fs: int = None) -> List:
+
+ def convert_to_wav(input_path, output_path):
+ from pydub import AudioSegment
+ try:
+ audio = AudioSegment.from_mp3(input_path)
+ audio.export(output_path, format="wav")
+ print("闊抽鏂囦欢涓簃p3鏍煎紡锛屽凡杞崲涓簑av鏍煎紡")
+
+ except Exception as e:
+ print(f"杞崲澶辫触:{e}")
+
def load_wav(path: str) -> np.ndarray:
+ if not path.lower().endswith('.wav'):
+ import os
+ input_path = path
+ path = os.path.splitext(path)[0]+'.wav'
+ convert_to_wav(input_path,path) #灏唌p3鏍煎紡杞崲鎴恮av鏍煎紡
+
waveform, _ = librosa.load(path, sr=fs)
return waveform
@@ -215,6 +230,10 @@
feats, feats_len = [], []
for waveform in waveform_list:
speech, _ = self.frontend.fbank(waveform)
+
+ if speech is None or speech.size == 0:
+ print("detected speech size {speech.size}")
+ raise ValueError("Empty speech detected, skipping this waveform.")
feat, feat_len = self.frontend.lfr_cmvn(speech)
feats.append(feat)
feats_len.append(feat_len)
--
Gitblit v1.9.1