Hirotome/DRA-Ctrl

Update requirements.txt

Update requirements.txt

<fix> remove resume_download

<feat> lower RAM usage when loading models

<feat> unify hacked_lora_forward

<fix> remove pipe_lock

<fix> fix a thread bug

<fix> fix a gradio version bug

<fix> fix a single thread bug

<feat> single thread

<fix> add backend for imageio

<fix> fix .gitattributes

<fix> re-upload example images

<fix> change .gitattributes

<fix> stop using git lfs

<fix> romove tempfile in requirements.txt

<fix> add some package requirements

<feat> add example images in assets

<feat> optimize output format and add examples

<fix> fix a model loading bug.

<feat> add diffusion inference steps control.

<fix> using whl for installing flash-attn.

<feat> using app.sh.

<feat> ZeroGPU only supports gradio sdk.

<feat> using docker sdk.

<feat> support changing models.

<feat> change output to gallery.

<fix> add tp ip attn enhancement control in forward.

<fix> fix a bug.

<fix> move to gpu earlier.

<fix> low cpu memory usage.

<fix> app using only one thread.

<fix> only leave lora weights loading in process_image_and_text.

<fix> fix some bugs.

<fix> fix a bug.

<fix> fix a bug.

<fix> move transformer init to process_image_and_text.

<fix> init base models at the beginning.

<fix> fix a syntax bug.

<fix> fix some bugs in app.py.

<fix> seperate model init and inference.

<fix> change torch and torchvision verisons to ZeroGPU-compatible versions.

<fix> unchange README.md.

<fix> change README.md.

<feat> add other links.

<fix> import spaces in app.py.

<feat> complete app.py.

<fix> fix create_app in app.py.

<fix> move flash_attn installation to app.py.

<fix> change flash_attn requirements.