tqchen commented on PR #16588: URL: https://github.com/apache/tvm/pull/16588#issuecomment-1955244162
I still think in this case the overall constraining the behavior of ConstIntBound is more predictable and readable. My main worry is we open a flood gate of introducing many recursive rewrite patterns to ConstIntBound itself. - Do you think if it is possible to introduce another subalayzer for this kind of prove given they are specific to LoRA, or have a stronger proof strength - In all cases, we should take a close look at the proves and make sure they are correct, since it is at center of what we are doing - If in the end we end up keep things in ConstIntBound, I think we should have a comment saying that this was really an exceptional case, and future changes to `ConstIntBound` should only be made in exceptional cases -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
