Skip to content

CVxTz/celery_ml_deploy

Repository files navigation

celery_ml_deploy

Sources

Some parts of the code are copied as is from https://huggingface.co/microsoft/Florence-2-base-ft/tree/main

This is done to solve a bug of transformers where the code cannot run without a GPU due to forceful import of flash_attn when the class definition is loaded with trust_remote_code=True.

Local Env Setup

CPU

pip install -r requirements.txt 
pip install -r app/requirements.txt
pip install -r model/requirements.txt --extra-index-url https://download.pytorch.org/whl/cpu

GPU (TODO: add support for GPU inference)

Docker

docker-compose up

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages