Revert "Fix parse matmulinteger 611" #2933
Closed
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
Originally we were using dot and converting the input for MatMulinteger() as it was parsed in. As a result we were inserting dot() instead of quant_dot into the model. I've added a change to check for similar data types and then add a shift to an input where there's a mismatch of int8 and uint8 on input. This is usually seen when dynamicquantizelinear is used as the input as it inserts a uint8 as the zero point which in some models (like gpt2) is operated on.
Handle tidy complexity error and break up functionality into seperate calls
Not needed and its flagged by our parse tests. Quant dot should by default output int32 for the accumulated output
- Add additional checks for input on supported types - Simplify add_int8_shit to check for not same supported 8-bit args - Allow for pass through for bias for quants -Remove extra layer of ifs for broadcast dims + dot/quant_dot output
…se_matmul Handle the following cases during parse - one bias (with broadcasts) - two bias (with broadcasts) - mismatch between zp and input - unsupported type Added cases to gen_onnx for error input
Fix parse MatMulinteger
|
do not merge, this will be deleted by me |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## release/rocm-rel-6.1 #2933 +/- ##
=====================================================
Coverage 91.46% 91.47%
=====================================================
Files 468 468
Lines 17542 17519 -23
=====================================================
- Hits 16045 16025 -20
+ Misses 1497 1494 -3 ☔ View full report in Codecov by Sentry. |
causten
force-pushed
the
release/rocm-rel-6.1
branch
from
c6f4a18 to
05a7707
Compare
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.
Reverts #2922