Implement aten::_softmax | feat(torchlib)
#1024
Merged
Conversation
This file contains hidden or 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
justinchuby
added
the
module: torchlib
Codecov Report
@@ Coverage Diff @@
## main #1024 +/- ##
==========================================
- Coverage 77.51% 77.50% -0.01%
==========================================
Files 114 114
Lines 14309 14335 +26
Branches 1521 1525 +4
==========================================
+ Hits 11091 11110 +19
- Misses 2854 2859 +5
- Partials 364 366 +2
|
fatcat-z
reviewed
Aug 25, 2023
|
|
||
| @torch_op("aten::_softmax", trace_only=True) | ||
| def aten__softmax( | ||
| self: TFloatHighPrecision, dim: int, half_to_float: bool |
There was a problem hiding this comment.
What does these 2 _softmax() functions has different dtypes for self?
There was a problem hiding this comment.
This is because we want to cast float16 to float32 when half_to_float is true. Since we have no idea what the input dtype will be within the function body, we rely on the dispatcher to pick the function that already limits the input dtype it can accept to know we are dealing with float16 types.
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
fatcat-z
reviewed
Aug 25, 2023
Test Results 18 files ± 0 18 suites ±0 1h 18m 35s ⏱️ - 9m 21s For more details on these failures and errors, see this check. Results for commit a1368e4. ± Comparison against base commit 74c068c. This pull request removes 266 and adds 277 tests. Note that renamed tests count towards both.♻️ This comment has been updated with latest results. |
fatcat-z
approved these changes
Aug 29, 2023
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Outdated suggestions cannot be applied.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.
Move
aten_special_softmax->aten_softmax, support alias and when dtype input is None. Implement_softmaxusingsoftmax.An alternative would be to create a decomp in pytorch from _softmax to softmax. But since pytorch optimization tends to create decomp to private ops (and not the other way around) in different optimizations steps as @BowenBao pointed out, we still need to support _softmax here.