heimoshuiyu/faster-whisper
1
0
Fork 0
Code Issues Pull Requests 1 Packages Projects Releases Wiki Activity

Force the garbage collector to run after decoding the audio with PyAV (#448)

Browse Source
This commit is contained in:
Guillaume Klein
2023-09-01 15:25:13 +02:00
committed by GitHub
parent e87fbf8a49
commit 5871858a5f
1 changed files with 6 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
faster_whisper/audio.py
View File

@@ -6,6 +6,7 @@ system dependencies. FFmpeg does not need to be installed on the system.
However, the API is quite low-level so we need to manipulate audio frames directly. However, the API is quite low-level so we need to manipulate audio frames directly.
""" """
import gc
import io import io
import itertools import itertools
@@ -53,6 +54,11 @@ def decode_audio(
dtype = array.dtype dtype = array.dtype
raw_buffer.write(array) raw_buffer.write(array)
# It appears that some objects related to the resampler are not freed
# unless the garbage collector is manually run.
del resampler
gc.collect()
audio = np.frombuffer(raw_buffer.getbuffer(), dtype=dtype) audio = np.frombuffer(raw_buffer.getbuffer(), dtype=dtype)
# Convert s16 back to f32. # Convert s16 back to f32.