From cbe2ea7e07cbf364827bd89cefc42b3f643ea3be Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期一, 18 三月 2024 23:59:09 +0800
Subject: [PATCH] paraformer streaming bugfix
---
funasr/utils/load_utils.py | 68 +++++++++++++++++++++++++++++++--
1 files changed, 63 insertions(+), 5 deletions(-)
diff --git a/funasr/utils/load_utils.py b/funasr/utils/load_utils.py
index 87412bd..6f12f55 100644
--- a/funasr/utils/load_utils.py
+++ b/funasr/utils/load_utils.py
@@ -14,7 +14,24 @@
except:
print("urllib is not installed, if you infer from url, please install it first.")
import pdb
+import subprocess
+from subprocess import CalledProcessError, run
+def is_ffmpeg_installed():
+ try:
+ output = subprocess.check_output(['ffmpeg', '-version'], stderr=subprocess.STDOUT)
+ return 'ffmpeg version' in output.decode('utf-8')
+ except (subprocess.CalledProcessError, FileNotFoundError):
+ return False
+
+use_ffmpeg=False
+if is_ffmpeg_installed():
+ use_ffmpeg = True
+else:
+ 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):
if isinstance(data_or_path_or_list, (list, tuple)):
@@ -31,11 +48,16 @@
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 data_type is None or data_type == "sound":
- 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)
+ 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)
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
@@ -88,8 +110,6 @@
return array
def extract_fbank(data, data_len = None, data_type: str="sound", frontend=None, **kwargs):
- # import pdb;
- # pdb.set_trace()
if isinstance(data, np.ndarray):
data = torch.from_numpy(data)
if len(data.shape) < 2:
@@ -114,3 +134,41 @@
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
+
+ Parameters
+ ----------
+ file: str
+ The audio file to open
+
+ sr: int
+ The sample rate to resample the audio if necessary
+
+ Returns
+ -------
+ A NumPy array containing the audio waveform, in float32 dtype.
+ """
+
+ # This launches a subprocess to decode audio while down-mixing
+ # and resampling as necessary. Requires the ffmpeg CLI in PATH.
+ # fmt: off
+ cmd = [
+ "ffmpeg",
+ "-nostdin",
+ "-threads", "0",
+ "-i", file,
+ "-f", "s16le",
+ "-ac", "1",
+ "-acodec", "pcm_s16le",
+ "-ar", str(sr),
+ "-"
+ ]
+ # fmt: on
+ try:
+ out = run(cmd, capture_output=True, check=True).stdout
+ except CalledProcessError as e:
+ raise RuntimeError(f"Failed to load audio: {e.stderr.decode()}") from e
+
+ return np.frombuffer(out, np.int16).flatten().astype(np.float32) / 32768.0
--
Gitblit v1.9.1