Bring in improvements to original GFPGAN

[cuuupid]

No description provided.

[OmlineEditor]

tell me have you added a new version to the replicate website: https://replicate.com/tencentarc/gfpgan/versions
in the new version on the site, what exactly have you changed? just what is here in this pull request?

[cuuupid]

Hey @OmlineEditor this repo is a fork of GFPGAN as it appeared on Replicate with a couple of tweaks for remote image URLs. This is what is currently live on Replicate.

This PR, once merged, would bring in changes to GFPGAN made to the additional repository (changes that were not pushed to Replicate).

[OmlineEditor]

then later all the improvements that you make will appear on the site replicate.com?

[OmlineEditor]

I have a request for you, can you change the launch command to process the image?
There is another popular model, but when processing images, “CUDA out of memory” errors often occur:
https://replicate.com/jingyunliang/swinir

This comment is indicated in the description of the program:
# 003 Real-World Image Super-Resolution (use --tile 400 if you run out-of-memory)

I see you were able to make a new version not for your program and perhaps you can for another.
can you change the command to run the program to reduce the number of errors?