Monitored training

@tiago.pereira this is a WIP but you are welcome to review when you have the time.

added 2 commits

• 78e36f6d - fix example
• daea0805 - use the dataset API and remove the un-needed extra

Compare with previous version

Well, I was writing this before our conversation, so I will finish it.... ::non-potable_water: ...

Man, this is super simple (specially the train script), thanks for the contrib.

Two things. You remember the thing I mentioned about the summaries ?? I think I can create a hook to gather the stuff I want (some parts of the loss). If I have a different kind of operation and I want to keep track of some stuff, I could just write a hook for it.

The other thing, we could hook the verify.py in the evaluation script.

added 1 commit

• f66693fa - Simplify scripts using tf.estimator API

Compare with previous version

added 1 commit

• a33eb9db - model_dir is important for evaluation

Compare with previous version

added 1 commit

• a154ac22 - remove template variables

Compare with previous version

added 1 commit

• a49cabae - Add an mnist example for estimator scripts. Fixed some bugs

Compare with previous version

unmarked as a Work In Progress

assigned to @tiago.pereira

@tiago.pereira this is ready.

Cool, thanks, I will need one hour to finish my stuff. I will merge it to the master. Then I will apply this MR and remove some stuff from the current master.

Rebasing

added 15 commits

• a49cabae...205ab9f4 - 6 commits from branch master
• 52023e4a - Move code from bob.dap to here
• 2956f8ea - autopep8
• 1fb310c4 - Improve scripts
• 0c6cdd35 - fix example
• 812e233b - use the dataset API and remove the un-needed extra
• 6524d975 - Simplify scripts using tf.estimator API
• a95b53f8 - model_dir is important for evaluation
• 5c4af81a - remove template variables
• 23d36b81 - Add an mnist example for estimator scripts. Fixed some bugs

Compare with previous version

merged

mentioned in commit 882d1245