From d80ac2fd2df4e7fb8a28acfa512bb11472b5cc99 Mon Sep 17 00:00:00 2001
From: liugz18 <57401541+liugz18@users.noreply.github.com>
Date: 星期四, 18 七月 2024 21:34:55 +0800
Subject: [PATCH] Rename 'res' in line 514 to avoid with naming conflict with line 365
---
funasr/utils/load_utils.py | 112 +++++++++++++++++++++++++++++++++++++++-----------------
1 files changed, 78 insertions(+), 34 deletions(-)
diff --git a/funasr/utils/load_utils.py b/funasr/utils/load_utils.py
index 4849408..346a00b 100644
--- a/funasr/utils/load_utils.py
+++ b/funasr/utils/load_utils.py
@@ -9,6 +9,7 @@
import time
import logging
from torch.nn.utils.rnn import pad_sequence
+
try:
from funasr.download.file import download_from_url
except:
@@ -17,53 +18,93 @@
import subprocess
from subprocess import CalledProcessError, run
+
def is_ffmpeg_installed():
try:
- # 灏濊瘯杩愯ffmpeg鍛戒护骞惰幏鍙栧叾鐗堟湰淇℃伅
- output = subprocess.check_output(['ffmpeg', '-version'], stderr=subprocess.STDOUT)
- return 'ffmpeg version' in output.decode('utf-8')
+ output = subprocess.check_output(["ffmpeg", "-version"], stderr=subprocess.STDOUT)
+ return "ffmpeg version" in output.decode("utf-8")
except (subprocess.CalledProcessError, FileNotFoundError):
- # 鑻ヨ繍琛宖fmpeg鍛戒护澶辫触锛屽垯璁や负ffmpeg鏈畨瑁�
return False
-
-use_ffmpeg=False
+
+
+use_ffmpeg = False
if is_ffmpeg_installed():
use_ffmpeg = True
else:
- print("Notice: ffmpeg is not installed. torchaudio is used to load audio")
+ print(
+ "Notice: ffmpeg is not installed. torchaudio is used to load audio\n"
+ "If you want to use ffmpeg backend to load audio, please install it by:"
+ "\n\tsudo apt install ffmpeg # ubuntu"
+ "\n\t# brew install ffmpeg # mac"
+ )
-def load_audio_text_image_video(data_or_path_or_list, fs: int = 16000, audio_fs: int = 16000, data_type="sound", tokenizer=None, **kwargs):
+
+def load_audio_text_image_video(
+ data_or_path_or_list,
+ fs: int = 16000,
+ audio_fs: int = 16000,
+ data_type="sound",
+ tokenizer=None,
+ **kwargs,
+):
if isinstance(data_or_path_or_list, (list, tuple)):
if data_type is not None and isinstance(data_type, (list, tuple)):
data_types = [data_type] * len(data_or_path_or_list)
data_or_path_or_list_ret = [[] for d in data_type]
- for i, (data_type_i, data_or_path_or_list_i) in enumerate(zip(data_types, data_or_path_or_list)):
- for j, (data_type_j, data_or_path_or_list_j) in enumerate(zip(data_type_i, data_or_path_or_list_i)):
- data_or_path_or_list_j = load_audio_text_image_video(data_or_path_or_list_j, fs=fs, audio_fs=audio_fs, data_type=data_type_j, tokenizer=tokenizer, **kwargs)
+ for i, (data_type_i, data_or_path_or_list_i) in enumerate(
+ zip(data_types, data_or_path_or_list)
+ ):
+ for j, (data_type_j, data_or_path_or_list_j) in enumerate(
+ zip(data_type_i, data_or_path_or_list_i)
+ ):
+ data_or_path_or_list_j = load_audio_text_image_video(
+ data_or_path_or_list_j,
+ fs=fs,
+ audio_fs=audio_fs,
+ data_type=data_type_j,
+ tokenizer=tokenizer,
+ **kwargs,
+ )
data_or_path_or_list_ret[j].append(data_or_path_or_list_j)
return data_or_path_or_list_ret
else:
- return [load_audio_text_image_video(audio, fs=fs, audio_fs=audio_fs, data_type=data_type, **kwargs) for audio in data_or_path_or_list]
- if isinstance(data_or_path_or_list, str) and data_or_path_or_list.startswith('http'): # download url to local file
+ return [
+ load_audio_text_image_video(
+ audio, fs=fs, audio_fs=audio_fs, data_type=data_type, **kwargs
+ )
+ for audio in data_or_path_or_list
+ ]
+ if isinstance(data_or_path_or_list, str) and data_or_path_or_list.startswith(
+ "http"
+ ): # download url to local file
data_or_path_or_list = download_from_url(data_or_path_or_list)
- if isinstance(data_or_path_or_list, str) and os.path.exists(data_or_path_or_list): # local file
+ if isinstance(data_or_path_or_list, str) and os.path.exists(data_or_path_or_list): # local file
if data_type is None or data_type == "sound":
- if use_ffmpeg:
- data_or_path_or_list = _load_audio_ffmpeg(data_or_path_or_list, sr=fs)
- data_or_path_or_list = torch.from_numpy(data_or_path_or_list).squeeze() # [n_samples,]
- else:
+ # if use_ffmpeg:
+ # data_or_path_or_list = _load_audio_ffmpeg(data_or_path_or_list, sr=fs)
+ # data_or_path_or_list = torch.from_numpy(data_or_path_or_list).squeeze() # [n_samples,]
+ # else:
+ # data_or_path_or_list, audio_fs = torchaudio.load(data_or_path_or_list)
+ # if kwargs.get("reduce_channels", True):
+ # data_or_path_or_list = data_or_path_or_list.mean(0)
+ try:
data_or_path_or_list, audio_fs = torchaudio.load(data_or_path_or_list)
if kwargs.get("reduce_channels", True):
data_or_path_or_list = data_or_path_or_list.mean(0)
+ except:
+ data_or_path_or_list = _load_audio_ffmpeg(data_or_path_or_list, sr=fs)
+ data_or_path_or_list = torch.from_numpy(
+ data_or_path_or_list
+ ).squeeze() # [n_samples,]
elif data_type == "text" and tokenizer is not None:
data_or_path_or_list = tokenizer.encode(data_or_path_or_list)
- elif data_type == "image": # undo
+ elif data_type == "image": # undo
pass
- elif data_type == "video": # undo
+ elif data_type == "video": # undo
pass
-
+
# if data_in is a file or url, set is_final=True
if "cache" in kwargs:
kwargs["cache"]["is_final"] = True
@@ -73,16 +114,16 @@
elif isinstance(data_or_path_or_list, np.ndarray): # audio sample point
data_or_path_or_list = torch.from_numpy(data_or_path_or_list).squeeze() # [n_samples,]
elif isinstance(data_or_path_or_list, str) and data_type == "kaldi_ark":
- data_mat = kaldiio.load_mat(data_or_path_or_list)
+ data_mat = kaldiio.load_mat(data_or_path_or_list)
if isinstance(data_mat, tuple):
audio_fs, mat = data_mat
else:
mat = data_mat
- if mat.dtype == 'int16' or mat.dtype == 'int32':
+ if mat.dtype == "int16" or mat.dtype == "int32":
mat = mat.astype(np.float64)
mat = mat / 32768
- if mat.ndim ==2:
- mat = mat[:,0]
+ if mat.ndim == 2:
+ mat = mat[:, 0]
data_or_path_or_list = mat
else:
pass
@@ -93,30 +134,32 @@
data_or_path_or_list = resampler(data_or_path_or_list[None, :])[0, :]
return data_or_path_or_list
+
def load_bytes(input):
middle_data = np.frombuffer(input, dtype=np.int16)
middle_data = np.asarray(middle_data)
- if middle_data.dtype.kind not in 'iu':
+ if middle_data.dtype.kind not in "iu":
raise TypeError("'middle_data' must be an array of integers")
- dtype = np.dtype('float32')
- if dtype.kind != 'f':
+ dtype = np.dtype("float32")
+ if dtype.kind != "f":
raise TypeError("'dtype' must be a floating point type")
-
+
i = np.iinfo(middle_data.dtype)
abs_max = 2 ** (i.bits - 1)
offset = i.min + abs_max
array = np.frombuffer((middle_data.astype(dtype) - offset) / abs_max, dtype=np.float32)
return array
-def extract_fbank(data, data_len = None, data_type: str="sound", frontend=None, **kwargs):
+
+def extract_fbank(data, data_len=None, data_type: str = "sound", frontend=None, **kwargs):
if isinstance(data, np.ndarray):
data = torch.from_numpy(data)
if len(data.shape) < 2:
- data = data[None, :] # data: [batch, N]
+ data = data[None, :] # data: [batch, N]
data_len = [data.shape[1]] if data_len is None else data_len
elif isinstance(data, torch.Tensor):
if len(data.shape) < 2:
- data = data[None, :] # data: [batch, N]
+ data = data[None, :] # data: [batch, N]
data_len = [data.shape[1]] if data_len is None else data_len
elif isinstance(data, (list, tuple)):
data_list, data_len = [], []
@@ -125,14 +168,15 @@
data_i = torch.from_numpy(data_i)
data_list.append(data_i)
data_len.append(data_i.shape[0])
- data = pad_sequence(data_list, batch_first=True) # data: [batch, N]
+ data = pad_sequence(data_list, batch_first=True) # data: [batch, N]
data, data_len = frontend(data, data_len, **kwargs)
-
+
if isinstance(data_len, (list, tuple)):
data_len = torch.tensor([data_len])
return data.to(torch.float32), data_len.to(torch.int32)
+
def _load_audio_ffmpeg(file: str, sr: int = 16000):
"""
Open an audio file and read as mono waveform, resampling as necessary
--
Gitblit v1.9.1