Fix bug in integer division #345
Conversation
Summary: In certain cases, when inputs are integers, division will statistically round rather than converting to floating point division. This happens when the encoder scale is set to 1, which will happen after a comparison operator (e.g. sign() etc.) I found this in the linear_svm example during evaluation as we compute `accuracy = correct.add(1).div(2).mean()`. The `div(2)` will round because the input encoder has scale 1 due to the sign in the SVM code. This diff corrects the issue and adds this edge case to testing. Differential Revision: D33686267 fbshipit-source-id: 06da5d3986120d8bf64b2db438586fcf3e4d9a6f
facebook-github-bot
added
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D33686267 |
|
Hi @knottb! Thank you for your pull request. We require contributors to sign our Contributor License Agreement, and yours needs attention. You currently have a record in our system, but the CLA is no longer valid, and will need to be resubmitted. ProcessIn order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA. Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with If you have received this in error or have any questions, please contact us at [email protected]. Thanks! |
Summary:
In certain cases, when inputs are integers, division will statistically round rather than converting to floating point division. This happens when the encoder scale is set to 1, which will happen after a comparison operator (e.g. sign() etc.)
I found this in the linear_svm example during evaluation as we compute
accuracy = correct.add(1).div(2).mean(). Thediv(2)will round because the input encoder has scale 1 due to the sign in the SVM code.This diff corrects the issue and adds this edge case to testing.
Differential Revision: D33686267