Decompose addmm with Gemm | feat(torchlib) #1111
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
justinchuby
commented
Oct 24, 2023
justinchuby
commented
Oct 24, 2023
Codecov Report
@@ Coverage Diff @@
## main #1111 +/- ##
==========================================
+ Coverage 78.13% 78.16% +0.02%
==========================================
Files 117 117
Lines 14954 14966 +12
Branches 1585 1586 +1
==========================================
+ Hits 11685 11698 +13
Misses 2900 2900
+ Partials 369 368 -1
|
|
|
BowenBao
reviewed
Oct 24, 2023
| @@ -232,6 +249,29 @@ def aten_addmm( | |||
| return op.Add(scaled_self, scaled_mat1_mat2) | |||
|
|
|||
|
|
|||
| @torch_op("aten::addmm") | |||
| def aten_addmm_gemm( | |||
| self: FLOAT, mat1: FLOAT, mat2: FLOAT, beta: float = 1.0, alpha: float = 1.0 | |||
There was a problem hiding this comment.
Why only do 'gemm' variant on float inputs?
There was a problem hiding this comment.
ORT implements Gemm for float inputs only. So I set it accordingly for practicality
There was a problem hiding this comment.
Changed to TFloat because it supports float16 as well
Test Results 18 files ± 0 18 suites ±0 1h 45m 59s ⏱️ + 21m 35s For more details on these failures and errors, see this check. Results for commit b448bc0. ± Comparison against base commit 9fb0a7d. This pull request removes 520 and adds 539 tests. Note that renamed tests count towards both.This pull request skips 8 and un-skips 20 tests.♻️ This comment has been updated with latest results. |
Yeah Gemm seems to be more specialized for that case. Backends should just do the right thing for matmul imo |
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.
Decompose addmm with Gemm by creating a special variant for
FLOATand conditionally check for the ranks if the input tensors. The if branch is expected to be folded away by constant folding passes.I have not found other instances where Gemm is used in the torch.onnx exporter.
Fixes #1089
cc @baijumeswani