You can also try Catalyst for more production-ready research ;)

Catalyst — @Pytorch framework for Deep Learning research and development. You get a training loop with metrics, model checkpointing, advanced logging and distributed training support without the boilerplate.

Do you want gentle introduction?
https://colab.research.google.com/drive/1T5G_OVLYHq483l80ikabXRsx_WB3rwPf?usp=sharing

Minimal examples for ML, CV, AE, VAE, GANs… with distributed training supprot?
https://github.com/catalyst-team/catalyst#minimal-examples

Documentation?
https://catalyst-team.github.io/catalyst/api/core.html#experiment

Guides?
https://catalyst-team.github.io/catalyst/info/distributed.html

Extra production ready pipelines?
https://github.com/catalyst-team/classification
https://github.com/catalyst-team/segmentation
https://github.com/catalyst-team/detection

Break the cycle — use the Catalyst!