JeffreyCA

Hi, can you see if installing this Tensorflow package solves the issue? I got it from the link you provided. https://snapshots.linaro.org/ldcg/python/tensorflow-manylinux/43/tensorflow-aarch64/tensorflow_aarch64-2.5.3-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl

Sorry for the late reply, you can set the `APP_HOST` environment variable or edit `ALLOWED_HOSTS` in `settings.py`/`settings_docker.py`

Is this when using Spleeter? Can you login to the celery-slow container and delete the `/webapp/pretrained_models/4stems` folder?

Can you share logs from the Celery process? If you're using Docker, what does this return? ``` docker exec -it spleeter-web-celery-slow-1 ls -la /webapp/pretrained_models ```

I think this is similar: https://github.com/deezer/spleeter/issues/306 You may need to download and extract https://github.com/deezer/spleeter/releases/download/v1.4.0/4stems.tar.gz yourself and copy it to `/webapp/pretrained_models`, not sure why it's not downloading the model automatically.

Hi, sorry for the late reply. Could you clarify the issue, and provide logs if possible?

Unfortunately due to the number of Python dependencies on different separation models, it creates a situation where certain subdependencies have to be on the lowest common version because of dependency...

Thanks for reorganizing! I'd like to keep the existing behaviour with `docker-compose.dev.yml` where a frontend service is running `npm run dev` (webpack in watch mode) and has a shared volume...

Yeah that's fine

No immediate plans but a good idea. Most of the strings that require localization are on the React/frontend side of things