Merge pull request #40 from florath/fix-MLP-for-non-gpu

Merge pull request #39 from florath/tensorflow1.7

Fix MLP network that it can be used with use_gpu=False.

Adds tensorflow 1.7 compatibility

add soft_allowment to not use all gpu for small network

n_repeat=1 only works

Merge pull request #28 from camallenlloyd/master

Update TF and gym compatibility

add tf.global_variables_initializer. fix #21

dummy to dummy

update TF compatability and add option use_cumulated_reward

Merge pull request #13 from dillonalaird/return-cumulated-reward

Merge pull request #14 from dillonalaird/play-train-false

is_training set to false for play

returns cumulated reward for each step