awslabs / keras-apache-mxnet

[DEPRECATED] Amazon Deep Learning's Keras with Apache MXNet support

https://github.com/awslabs/keras-apache-mxnet/wiki

Other

290 stars 65 forks source link

Prepare for keras-mxnet 2.2 release #117

Closed sandeep-krishnamurthy closed 6 years ago

sandeep-krishnamurthy commented 6 years ago

Changes

Pull and rebase keras-team/keras v2.2.0 release
Fix all changes on the backend, tests, examples

Testing Done:

All UTs / ITs are passing
All examples are running
Testing on GPU machines - Default usage of 1st GPU and multi_gpu_model for Sequential and Model
Multi GPU models - Sequential / Model

@roywei @kalyc

sandeep-krishnamurthy commented 6 years ago

updated the issue with all the changes - https://github.com/awslabs/keras-apache-mxnet/issues/92

sandeep-krishnamurthy commented 6 years ago

Thanks @kalyc your feedback was very helpful. Fixed it.

sandeep-krishnamurthy commented 6 years ago

Fixed class_weight ignore issue.
Fixed accuracy issue with custom loss.
Added warning message for a workaround if users are using custom loss and slice operation. Linked to issue here - https://github.com/awslabs/keras-apache-mxnet/issues/120
Updated Conv1D test

I would like to merge this PR to master prepare v 2.2 beta2 build that should be tested with latest MXNet 1.2.1 and make a release. We can then updated dev from master post release. Suggestions?

@roywei @kalyc