Skip to content

Enable assign operator for a ngraph, test=develop#17437

Merged
tensor-tang merged 2 commits intoPaddlePaddle:developfrom
mozga-intel:mozga-intel/assign_operator_ngraph
May 24, 2019
Merged

Enable assign operator for a ngraph, test=develop#17437
tensor-tang merged 2 commits intoPaddlePaddle:developfrom
mozga-intel:mozga-intel/assign_operator_ngraph

Conversation

@mozga-intel
Copy link
Contributor

@mozga-intel mozga-intel commented May 15, 2019

Enable the assign operator for the nGraph bridge. BERT model uses it.

Additionally, the corss_entropy test was updated, due to that CI's test fails.

@mozga-intel mozga-intel added this to the v1.5 for Intel milestone May 15, 2019
@mozga-intel mozga-intel requested a review from luotao1 May 15, 2019 23:23
@luotao1 luotao1 requested a review from tensor-tang May 16, 2019 03:33
@mozga-intel mozga-intel force-pushed the mozga-intel/assign_operator_ngraph branch 2 times, most recently from 8adfb6f to 22a222e Compare May 20, 2019 21:38
Copy link
Contributor

@tensor-tang tensor-tang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Additionally, the corss_entropy test was updated, due to that CI's test fails.

Could u give some more details about this ?

@mozga-intel
Copy link
Contributor Author

@tensor-tang The assign operator needs to support different shapes so any changes of xe operator are needed. The previous version of xe operator did not include these changes.

@tensor-tang
Copy link
Contributor

tensor-tang commented May 23, 2019

xe operator are needed

did not understand this

due to that CI's test fails.

Actually I mean what is the test fails you said?

@mozga-intel
Copy link
Contributor Author

mozga-intel commented May 23, 2019

@tensor-tang

  1. xe operator are needed - it means that I must have updated the cross_entoropy operator
  2. Actually, I mean what is the test fails you said? - cross_entropy_ngraph

Copy link
Contributor

@tensor-tang tensor-tang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@tensor-tang tensor-tang merged commit f2694e1 into PaddlePaddle:develop May 24, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants