bgawrych opened a new pull request #20323:
URL: https://github.com/apache/incubator-mxnet/pull/20323


   ## Description ##
   Fix for fusing requantize and dequantize into quantized op when it's outputs 
have different data types:
   ```
                    /---> Quantizable op
    Input ---> FC -|
                    \---> Non quantizable op
   ```
   
![image](https://user-images.githubusercontent.com/59644968/120325191-b54eb400-c2e7-11eb-90dc-270537206e91.png)
   
   Found while quantizing google_electra models
   
   Also this PR fixes invalid layer name requested for calibration - occurs 
when input to quantize node is variable
   
   ## Checklist ##
   ### Essentials ###
   - [x] PR's title starts with a category (e.g. [BUGFIX], [MODEL], [TUTORIAL], 
[FEATURE], [DOC], etc)
   - [x] Changes are complete (i.e. I finished coding on this PR)
   - [x] All changes have test coverage
   - [x] Code is well-documented
   
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to