roboflow / inference

A fast, easy-to-use, production-ready inference server for computer vision supporting deployment of many popular model architectures and fine-tuned models.

https://inference.roboflow.com

Other

1.12k stars 84 forks source link

Add florence 2 to inference, better LoRA support #486

Closed probicheaux closed 3 days ago

probicheaux commented 1 week ago

Description

Refactor paligemma model impl to be more general and add florence 2 based off the refactor

Type of change

Please delete options that are not relevant.

[ ] Bug fix (non-breaking change which fixes an issue)
[x] New feature (non-breaking change which adds functionality)
[ ] This change requires a documentation update

How has this change been tested, please provide a testcase or example of how you tested the change?

Tested with staging api response

Any specific deployment considerations

Docs

[ ] Docs updated? What were the changes:

probicheaux commented 1 week ago

sorry diff looks weird, but most of the content has been moved inference/models/paligemma/paligemma.py -> inference/models/transformers/transformers.py

probicheaux commented 3 days ago

@grzegorz-roboflow

I assume code loaded with import_class_from_file (transformer class and processor class) is reviewed by us if we are going to host this.

yah, this is currently only called on code that is uploaded to our core models bucket, which is only hosted by us and uploaded manually. I've reviewed every file in there (and they come from the official microsoft release of florence-v2, so nothing to worry about)

Gonna merge as it seems all comments have been addressed