reductionista opened a new pull request #524:
URL: https://github.com/apache/madlib/pull/524
We don't want to allow arbitrary custom callback functions, but we can make
a special exception for TensorBoard.
With this commit, users will be able to get realtime feedback on accuracy
and loss while training, as well as visualize the tensor graph of the keras
model they've submitted.
Usage:
1. Add callback to TensorBoard to fit_params:
```
callbacks=[TensorBoard(log_dir="/tmp/tensorflow/scalars")]
```
2. Start tensorboard before training or prediction with MADlib deep
learning module:
```
tensorboard --logdir='/tmp/tensorflow/scalars'
To quickly enable it for an entire model selection table:
```
update mst_table set fit_params = (fitparams || ',
callbacks=[TensorBoard(log_dir="/tmp/tensorflow/scalars")]');
```
We should look over it carefully to make sure this doesn't add significant
security risk before merging.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
