Future changes to boost computational efficiency

Great PR @ajlee21. Combination of nitpicking coding style-wise and some confusion towards the purpose of some of your scripts/notebooks. After talking in person I have two practical suggestions to make this code run smoother:

Would suggest uploading your data onto the pmacs cluster. Since it is a cluster, it should provide access to greater memory than your personal computer. Plus it would allow for GPU availability, which in turn would speed up your computation.
Would recommend changing frameworks to pytorch, since it contains a dynamic computational graph compared to tensorflow's static graph. (i.e. Memory allocation is a lot smoother and the most amount of memory one would need is equal to batch size you feed the network.) Obviously this option would take more work, but something to think about as this project progresses.

from https://github.com/greenelab/LINCS_latent_space/pull/1#pullrequestreview-192310756