Just want to take the temperature of the community at this point in time. What are you using for research and production right now, tensorflow or pytorch? What is your experiences with speed of development for writing papers, code quality and eventual release to a production environment?

And if you want to elaborate what more do you use in conjunction, git for versioning, tensorboard for monitoring and docker for environment or something completely different?

Personally I’ve been using pytorch for smaller experiments and my local environment but will start doing larger experiments and want to learn from you all.