From c644ac8f58895b9e29e9cfca79465fd2c0efaa5a Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期二, 21 十一月 2023 14:09:01 +0800
Subject: [PATCH] funasr v2 setup
---
funasr/utils/wav_utils.py | 6 +++---
1 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/funasr/utils/wav_utils.py b/funasr/utils/wav_utils.py
index bd067c2..8c2dc68 100644
--- a/funasr/utils/wav_utils.py
+++ b/funasr/utils/wav_utils.py
@@ -11,7 +11,7 @@
import numpy as np
import torch
import torchaudio
-import soundfile
+import librosa
import torchaudio.compliance.kaldi as kaldi
@@ -166,7 +166,7 @@
try:
waveform, audio_sr = torchaudio.load(wav_file)
except:
- waveform, audio_sr = soundfile.read(wav_file, dtype='float32')
+ waveform, audio_sr = librosa.load(wav_file, dtype='float32')
if waveform.ndim == 2:
waveform = waveform[:, 0]
waveform = torch.tensor(np.expand_dims(waveform, axis=0))
@@ -191,7 +191,7 @@
try:
waveform, sampling_rate = torchaudio.load(wav_path)
except:
- waveform, sampling_rate = soundfile.read(wav_path)
+ waveform, sampling_rate = librosa.load(wav_path)
waveform = torch.tensor(np.expand_dims(waveform, axis=0))
speech_length = (waveform.shape[1] / sampling_rate) * 1000.
n_frames = (waveform.shape[1] * 1000.0) / (sampling_rate * frontend_conf["frame_shift"] * frontend_conf["lfr_n"])
--
Gitblit v1.9.1