Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Conversion tool protobuffer -> flatbuffer #7460

Merged
merged 6 commits into from Apr 11, 2019

Conversation

Projects
None yet
3 participants
@yquemener
Copy link
Contributor

commented Apr 6, 2019

What changes were proposed in this pull request?

Added a nd4j-tools module that contains a small tool able to convert models from the protocol buffer format to the flat buffer format. Also includes a BERT pre-processor that removes the operations that prevents it from running with libnd4j. (As is done in BERTGraphTest)

This is useful in order to execute models with only the libnd4j C++ API, which can't import protocol buffer models.

Note that I did not add nd4j-tools in nd4j/pom.xml. That can be a totally optional build.

How was this patch tested?

I converted a densenet model and ran it through a libnd4j program, verified it produces the desired results.

yquemener added some commits Apr 6, 2019

Removed unnecessary execution tests, as there are better ones in the …
…main test suite now.

Added a junit test in order to easily start the converter from the IDE

@saudet saudet requested a review from AlexDBlack Apr 7, 2019

@AlexDBlack
Copy link
Member

left a comment

Let's put this here instead (under a "util" sub-folder).
https://github.com/deeplearning4j/deeplearning4j/tree/master/nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/imports/tensorflow

i.e., org.nd4j.imports.tensorflow.util.

I don't think we want to introduce a new module for just this.

yquemener added some commits Apr 9, 2019

@yquemener

This comment has been minimized.

Copy link
Contributor Author

commented Apr 9, 2019

And I clicked the "update branch" too hastily, thinking it would pull my latest commit, not masters' one. Hopefully it does not seem to have broken anything.

@saudet saudet requested a review from AlexDBlack Apr 11, 2019

@AlexDBlack
Copy link
Member

left a comment

LGTM (other than copyright header @saudet flagged)

@saudet

saudet approved these changes Apr 11, 2019

@saudet saudet merged commit c62e228 into deeplearning4j:master Apr 11, 2019

0 of 2 checks passed

Codacy/PR Quality Review Not up to standards. This pull request quality could be better.
Details
continuous-integration/jenkins/pr-head This commit is being built
Details
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.