tk1012 opened a new pull request #8942: URL: https://github.com/apache/tvm/pull/8942
This PR fixes a bug in convert_unpack(). This func currently convert `unpack(input, axis=0, num_packs=1)` into `squeeze(input, axis=None)`. However, this results in wrong results when `input` has multiple dimensions of size 1. For example, when `input`'s shape is `(1, 3, 4, 1)` as I added in the test code, a converted relay outputs a tensor which shape is `(3, 4)`. But, the correct output is `(3, 4, 1)`. To fix this problem, I made a change to always use `[unpack_axis]`. -- 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]
