Skip to content

sehoffmann/dmlcloud

Repository files navigation

dmlcloud

Flexibel, easy-to-use, opinionated

dmlcloud is a library for distributed training of deep learning models with torch. Its main aim is to do all these tiny little tedious things that everybody just copy pastes over and over again, while still giving you full control over the training loop and maximum flexibility.

Unlike other similar frameworks, such as lightning, dmcloud tries to add as little additional complexity and abstraction as possible. Instead, it is tailored towards a careful selected set of libraries and workflows and sticks with them.

About

Distributed torch training using horovod and slurm

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages