Add GroupNorm and LayerNorm onnx parsing #2242
Merged
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
Codecov Report
@@ Coverage Diff @@
## develop #2242 +/- ##
===========================================
+ Coverage 91.30% 91.33% +0.03%
===========================================
Files 436 438 +2
Lines 16345 16425 +80
===========================================
+ Hits 14923 15001 +78
- Misses 1422 1424 +2
|
attila-dusnoki-htec
force-pushed
the
onnx_missing_norms
branch
from
fd748bb to
39571f1
Compare
umangyadav
reviewed
Oct 10, 2023
Would be good to add that. But not sure about priority. We already have matcher that pattern matches primitives to LayerNorm. Perhaps it can be tweaked to match for group norm as well. |
attila-dusnoki-htec
force-pushed
the
onnx_missing_norms
branch
from
39571f1 to
59bb7b1
Compare
umangyadav
approved these changes
Oct 11, 2023
CharlieL7
requested changes
Oct 11, 2023
Comment on lines
+110
to
+122
| if(skipped_axes > 0) | ||
| { | ||
| auto x_dims = x_shape.lens(); | ||
| scale_bcast = info.add_instruction( | ||
| make_op("broadcast", {{"axis", skipped_axes}, {"out_lens", x_dims}}), scale); | ||
| if(not skip_bias) | ||
| { | ||
| bias_bcast = info.add_instruction( | ||
| make_op("broadcast", {{"axis", skipped_axes}, {"out_lens", x_dims}}), bias); | ||
| } | ||
| } |
There was a problem hiding this comment.
Note that this will not work if x_shape is dynamic. Probably outside of what we want to get done with this PR however. So should be fine.
attila-dusnoki-htec
force-pushed
the
onnx_missing_norms
branch
from
59bb7b1 to
7b39645
Compare
CharlieL7
approved these changes
Oct 13, 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Outdated suggestions cannot be applied.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.
This PR extends the normalization support with GroupNorm and LayerNorm.
Since they are very similar in implementation and usage, it is grouped in one PR.
I saw that there was already support for layernorm. This extends the parsing of the onnx node, but still uses built in primitives for calculation.
Similar support for groupnorm is not added yet. Is it needed? dnnl supports it. If so, i would recommend a follow-up PR instead of extending this.
Also MIOpen just introduced layernorm as a primitive, but it is in BETA.
Resolves migraphx-benchmark#94 and migraphx-benchmark#100