[ask] is it possible to copy weights from PyTorch model to complied AIT model

[YaoAstro]

I have a group of fine-tuned stable diffusion models using the same training process. After compiling one model into AIT format, is it possible to reuse the "selected" AIT graph for other models by directly copying weights over (as opposed to compiling all the other models)?

[khabinov]

Hi @YaoAstro,

Such functionality should be more or less working though not very well documented. However, you can refer to this test to get some idea https://github.com/facebookincubator/AITemplate/blob/32a846507b129003c26600b1bcbe98f93aebcefe/tests/unittest/compiler/test_constant_folding.py#L543.

cc @muchulee8

[YaoAstro]

Thanks for the pointer, @khabinov!