WhisperX serving on Mosec

WhisperX splits single audio input into segments to parallel inference, but this project ONLY focus on using the batch-infernecing enabled faster-whisper to serve multiple request.

Setup

Please follow https://github.com/m-bain/whisperX to install conda env, then

pip install -r requirements.txt

to install mosec framework and other packages.

Usage

python main.py

or

Build Docker Image

You can choose to use the one of the following method:

• Dockerfile
  • docker buildx build -t <docker_hub_user_name>/<image_name> --push .
• build.envd
  • envd build -f :whisper_serving --output type=image,name=docker.io/<docker_hub_user_name>/<image_name> --push

Test

python client.py