notes.md

python -m tensorflow.contrib.learn.python.learn.utils.inspect_checkpoint

Useful Links:

• Intro to training recurrent nets on text. Very simple into, basis on how first experiments were carried out.
• Similar to above, simplified lstm training

Experiment: use min for kl loss term (look up a good value)

Table:

type of input input -> output for both random samples + encoded reps of samples