After running the script gmm_svae_synth.py (around one and a half hours on a Mac desktop computer), the results seem not reasonable (running_gmm_2000.png). So I just wondering if there is anything wrong with the script?
Yeah I left the example code in a broken state. It might be less broken on the gmm-synth-better-plots branch, but I'm not sure. I've been busy with a job search but I hope to come back and fix things up at some point in the not-too-distant future.