From c568628130ac42ebeea8cf48fe926520a31ff511 Mon Sep 17 00:00:00 2001
From: 嘉渊 <wangjiaming.wjm@alibaba-inc.com>
Date: 星期二, 16 五月 2023 10:57:21 +0800
Subject: [PATCH] update repo
---
funasr/fileio/sound_scp.py | 7 ++++---
1 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/funasr/fileio/sound_scp.py b/funasr/fileio/sound_scp.py
index cec7cd9..7b35a35 100644
--- a/funasr/fileio/sound_scp.py
+++ b/funasr/fileio/sound_scp.py
@@ -1,7 +1,6 @@
import collections.abc
from pathlib import Path
from typing import Union
-from typing import Optional
import random
import numpy as np
@@ -9,6 +8,7 @@
import librosa
from typeguard import check_argument_types
+import torch
import torchaudio
from funasr.fileio.read_text import read_2column_text
@@ -36,7 +36,7 @@
always_2d: bool = False,
normalize: bool = False,
dest_sample_rate: int = 16000,
- speed_perturb: Optional[list, tuple] = None,
+ speed_perturb: Union[list, tuple] = None,
):
assert check_argument_types()
self.fname = fname
@@ -63,8 +63,9 @@
speed = random.choice(self.speed_perturb)
if speed != 1.0:
array, _ = torchaudio.sox_effects.apply_effects_tensor(
- array, rate,
+ torch.tensor(array).view(1, -1), rate,
[['speed', str(speed)], ['rate', str(rate)]])
+ array = array.view(-1).numpy()
return rate, array
--
Gitblit v1.9.1