Longer term plan for prototype/inductor

[jerryzh168]

Currently we have inductor passes in https://github.com/pytorch/ao/tree/main/torchao/prototype/inductor for pt2e quantization x86 inductor backend, probably want to move these to pt2e? I remember it's because inductor does not have an official API to register passes out of pytorch yet. cc @jansel @eellison

[Xia-Weiwen]

Hi @jerryzh168 If I understand correctly, the inductor passes in https://github.com/pytorch/ao/tree/main/torchao/prototype/inductor is not for pt2e quantization x86 backend. It's for optimization of LLM quantization on X86 CPU.
CC @Valentine233

[jerryzh168]

@Xia-Weiwen I saw there is int8 sdpa etc. these are using pt2e quant with X86InductorQuantizer right? https://github.com/pytorch/ao/pull/2215/files#diff-d96f56237c6ec522900d166495c242d9dd49139041553786b6f43f6d8f227aeb