masahi commented on pull request #8921: URL: https://github.com/apache/tvm/pull/8921#issuecomment-912803305
@lazycal That is an interesting suggestion. Layout annotation in TVM was introduced a long time ago, and I believe this is the simplest solution that works in most cases. Indeed, `F.conv2d(x, F.conv2d(y, z))` would break in the current system. What you suggested makes a lot of sense and I like `I0_I1_10i0` annotation you mentioned in the discuss post. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
