From 8167ecaa49108c420c16deabdc4bdca3b75cf286 Mon Sep 17 00:00:00 2001
From: 嘉渊 <wangjiaming.wjm@alibaba-inc.com>
Date: 星期五, 12 五月 2023 11:56:44 +0800
Subject: [PATCH] update repo
---
funasr/datasets/large_datasets/dataset.py | 3 ++-
funasr/fileio/sound_scp.py | 4 +++-
2 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/funasr/datasets/large_datasets/dataset.py b/funasr/datasets/large_datasets/dataset.py
index 33ed13a..748e4bf 100644
--- a/funasr/datasets/large_datasets/dataset.py
+++ b/funasr/datasets/large_datasets/dataset.py
@@ -135,7 +135,8 @@
speed = random.choice(self.speed_perturb)
if speed != 1.0:
mat, _ = torchaudio.sox_effects.apply_effects_tensor(
- mat, sampling_rate, [['speed', str(speed)], ['rate', str(sampling_rate)]])
+ torch.tensor(mat).view(1, -1), sampling_rate, [['speed', str(speed)], ['rate', str(sampling_rate)]])
+ mat = mat.view(-1).numpy()
sample_dict[data_name] = mat
sample_dict["sampling_rate"] = sampling_rate
if data_name == "speech":
diff --git a/funasr/fileio/sound_scp.py b/funasr/fileio/sound_scp.py
index 11d582c..fc2c3e3 100644
--- a/funasr/fileio/sound_scp.py
+++ b/funasr/fileio/sound_scp.py
@@ -8,6 +8,7 @@
import librosa
from typeguard import check_argument_types
+import torch
import torchaudio
from funasr.fileio.read_text import read_2column_text
@@ -62,8 +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