Lower AtenFull op #5781
Merged
Lower AtenFull op #5781
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
wonjoolee95
approved these changes
Nov 10, 2023
There was a problem hiding this comment.
Thanks for working on this, @danielvegamyhre!
- Do we want to fallback to the pytorch/aten CPU version of this op if the
layoutand/orpin_memoryvalues are not the default? I saw this pattern in other places in inaten_xla_type.cpp, and thetensor_methods::fullfunction in XLA doesn't support these parameters, so it seemed like the right thing to do, but I'm not sure.
Yes, this is the correct behavior. We can fallback to CPU for now for full like you did in this PR.
- I notice
fullwas not listed inxla_native_functions.yaml(as expected) but there IS an existing test for it intest_aten_xla_tensor_1.cppcalledTestFull, which seems to verify that we usexla::emptyfollowed byxla::fill_to create a tensor with the fill value. So I just changed this test to check that we are instead using the newfullmethod implemented in this PR. I confirmed the test is passing with these changes, and failing without the changes. Is this the correct thing to do here?
Good find -- so most likely what's happening here is that when full is not lowered, it's decomposed empty and fill. But since this PR lowers full, what you did here -- removing the counter checks for empty and fill -- is correct, thanks!
LGTM.
|
Awesome! Thank you @danielvegamyhre |
mbzomowski
pushed a commit
to mbzomowski-test-org/xla
that referenced
this pull request
Nov 16, 2023
* lower full * update test for full op * formatting
zpcore
pushed a commit
that referenced
this pull request
Nov 21, 2023
* lower full * update test for full op * formatting
lsy323
pushed a commit
to lsy323/xla
that referenced
this pull request
Nov 28, 2023
* lower full * update test for full op * formatting
chunnienc
pushed a commit
to chunnienc/xla
that referenced
this pull request
Dec 14, 2023
* lower full * update test for full op * formatting
golechwierowicz
pushed a commit
that referenced
this pull request
Jan 12, 2024
* lower full * update test for full op * formatting
bhavya01
pushed a commit
that referenced
this pull request
Apr 22, 2024
* lower full * update test for full op * formatting
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #5780
This is my first attempt at op lowering, and I have a couple questions:
Do we want to fallback to the pytorch/aten CPU version of this op if the
layoutand/orpin_memoryvalues are not the default? I saw this pattern in other places in inaten_xla_type.cpp, and thetensor_methods::fullfunction in XLA doesn't support these parameters, so it seemed like the right thing to do, but I'm not sure.I notice
fullwas not listed inxla_native_functions.yaml(as expected) but there IS an existing test for it intest_aten_xla_tensor_1.cppcalledTestFull, which seems to verify that we usexla::emptyfollowed byxla::fill_to create a tensor with the fill value. So I just changed this test to check that we are instead using the newfullmethod implemented in this PR. I confirmed the test is passing with these changes, and failing without the changes. Is this the correct thing to do here?