Closed abatilo closed 11 months ago
I received this response to this question:
We are not maintaining this library anymore, but the core algorithm we proposed in Alpa has been merged into the official XLA https://github.com/openxla/xla/tree/main/xla/hlo/experimental/auto_sharding. A Google team is maintaining it.
Thank you for sharing @0xc1c4da. Maybe maintainers can add that to the README and archive the project
The last merged commit was from a while ago and the supported version of Ray is lagging behind as well. This project seems hugely important for optimally training larger models so I'd be curious to know if development is still active or if this project is going stale.