Fix strided_slice legalization when input_dim % stride != 0
- Add padding to round an input dim up to the nearest multiple of the corresponding stride value
Change-Id: I1b9677c80e7ba704897637a3c44545d24cdae892
Hi @rsuderman, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
Hi @rdzhabarov, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
Hi @jpienaar, Can you please review this PR ? Thank you!
@jamwar01 can you rebase this patch to solve the conflict, so that we can proceed with review and merge? Thanks
Am I correct in thinking that this patch solves the same problem as another open patch, https://github.com/tensorflow/tensorflow/pull/61424?
@jamwar01 can you have a look on @ekalda's question and update the PR to solve conflicts if it is still the case? Thanks.
Hey @ekalda, yes it does look like both patches address the same issue -- good catch! Let me know how you would like to proceed.
Hi @jamwar01, thanks for looking again into the issue! I did some git archeology and it seems like the main logic of this patch was added with https://github.com/tensorflow/tensorflow/commit/0f1388eee0b7ceb3660bc9a0790ab2040538af06 (I'm not sure why) and the tests were added in https://github.com/tensorflow/tensorflow/pull/88530 just few weeks ago. So I think this PR and the duplicate can both be closed.
Ok thanks for digging into this!