Nandish Jayaram created MADLIB-1330:
---------------------------------------
Summary: DL: Do not compile model for predict and add GPU support
Key: MADLIB-1330
URL: https://issues.apache.org/jira/browse/MADLIB-1330
Project: Apache MADlib
Issue Type: Improvement
Components: Deep Learning
Reporter: Nandish Jayaram
Fix For: v1.16
In internal_keras_predict, we compile the model for every row. Starting from
keras 1.0.3, keras predict does not need a compiled model. This will make our
current predict code(which does not use SD) easier and hopefully slightly
faster even with CPU. See more details here
https://github.com/keras-team/keras/issues/3074
Acceptance:
1. Validate that removing the compile code does not impact the accuracy of the
results
2. Add new optional param named use_gpu in predict interface.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
