why need Uvicorn?
prprprus opened this issue · comments
Hi @tiangolo
As far as I understand, Uvicorn is a single process, which is very unfavorable for CPU-bound ML tasks. Why not just use gunicorn? If you add an extra layer of Uvicorn, it should increase performance consumption. It is this harmful to online prediction services. I look forward to your advice.
Please check the docs: https://github.com/tiangolo/uvicorn-gunicorn-machine-learning-docker#technical-details
Assuming the original issue was solved, it will be automatically closed now. But feel free to add more comments or create new issues.