Leaked semaphore warning on shutdown when using WhisperModel

Hi,

I'm using faster-whisper inside a container and noticing a warning on shutdown after running a transcription:

/usr/local/lib/python3.14/multiprocessing/resource_tracker.py:475: UserWarning: 
resource_tracker: There appear to be X leaked semaphore objects to clean up at shutdown

Minimal setup

• Python: 3.14
• faster-whisper:1.2.1
• Running inside a container
• Simple usage:

from faster_whisper import WhisperModel

model = WhisperModel("Systran/faster-whisper-base", device="auto", compute_type="auto")

segments, info = model.transcribe(audio_bytes)

What I observed

• The transcription works fine
• The warning appears only at process/container shutdown
• I tried explicitly cleaning up:
  • deleting the model (del model)
  • clearing internal references (model.model = None)
  • calling gc.collect()

…but the warning still appears.

Questions

1. Is this a known/common issue when using faster-whisper (or ctranslate2)?
2. Are there recommended ways to properly release resources or avoid this warning?
3. Is this something that needs to be handled at the library level, or is it expected/harmless?