Transcription models are currently automatically downloaded on request by whisper-service. However, for docker containers local files are cleared when a container is run, but it is not desirable to redownload models every time the container is changed. There should be a method to mount a directory so that models are stored directly on the host machine instead.
- Add script to whisper-service to download models specified in
device_configuration to some defined directory.
- Edit models to load from directory instead of automatically downloading model.
- Update README/Wiki to let user know to run script before running app (dev and prod)
Note: Depends on #12 but unrelated to #13 #14
Transcription models are currently automatically downloaded on request by whisper-service. However, for docker containers local files are cleared when a container is run, but it is not desirable to redownload models every time the container is changed. There should be a method to mount a directory so that models are stored directly on the host machine instead.
device_configurationto some defined directory.Note: Depends on #12 but unrelated to #13 #14