kevinthesun edited a comment on issue #6594: URL: https://github.com/apache/incubator-tvm/issues/6594#issuecomment-706474453
Hmmm, I think that PR is an optimization for some cases but is actually not correct since static dim is just a case of ```Any``` but not vise versa. The case provided in that PR only stands when ```False``` is fed with (Any, 1) tensor at runtime. In TF frontend there are lots of cases when we want to make expression as static as possible. My guessing is that ```False``` actually has shape (Any, 1) but somehow not fully optimized to be less dynamic. Usually such kind of optimization is done in the frontend. @lixiaoquan We should revert it? ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected]
