logicalclocks / maggy

Distribution transparent Machine Learning experiments on Apache Spark
https://maggy.ai
Apache License 2.0
89 stars 14 forks source link

Added ZeRO functionality, documentation, TypeHints #95

Closed amacati closed 3 years ago

amacati commented 3 years ago

Expanded PyTorch patching, added documentation and TypeHints. Integrates ZeRO.

Known issues: Datasets have to be even on all workers for training to succeed.

amacati commented 3 years ago

Added multiple notebooks, tensorboard functionality and addressed several small things.