Any particular reason int8 quantization is only supported on xl models?

[qingswu]

As the code checks, stable diffusion quantization only supports on SDXL models, does it work in SD1.4 or 1.5?

TensorRT/demo/Diffusion/utilities.py

Line 476 in 6d2fa4d

if args.int8 and not args.version.startswith('xl'):

[brb-nv]

Hi, thanks for checking here. Looks like a few things have changed since and support has been added for SD1.5 in TensorRT Model Optimizer here. Please check that out.

[qingswu]

Got it, thanks for the reference link!