Would it be okay to add a tqdm progressbar in main.py instead of logging every 2500 predictions to keep better track of progress?
I tested it in the Docker container and there weren’t any issues. I can make a pull request if this is okay.
@daniong Thanks for asking, and it’s a nice thought – we love tqdm and use it all the time. Unfortunately, while tqdm is great in the interactive case it seems to clutter up piped logs with many useless lines. If you find a neat way to get around this and test it locally using the entrypoint script we’d be happy to review a PR.