Amog Kamsetty comments

Results 55 comments of


                                            Amog Kamsetty

[WIP] Prototype splitting blocks to ensure requested read parallelism is always met

Closes https://github.com/ray-project/ray/issues/31501

[Templates] Unify the batch inference template with an existing Data example

Yep that’s right it will only run 1x, barring a few extra samples

[Data] `ray.data.from_huggingface` does not work as expected

Can you try again on master?

[Data][Docs] Remove "Data Representations (internal)"

seems like block is being referenced here `/ray/doc/source/data/examples/custom-datasource.rst:97: WARNING: undefined label: block-api`, and generally is part of the DeveloperAPI for creating new datasources.

[Data][Docs] Remove "Data Representations (internal)"

I think it's fine to keep under Data Representations. And we can downgrade it to developer API.

[Data] [no_early_kickoff] Change autoscaling logic to use free slots and input queue size

Do we have any existing actor pool autoscaling integration tests?

Not detecting GPU within trials when HPO

Hey @yinweisu that's right by default only the training worker processes reserve and use GPU, but the main process for each trial does not use GPU. This is because the...

Not detecting GPU within trials when HPO

Yes that would still work!

Not detecting GPU within trials when HPO

Hey @yinweisu the trainable function is just run once, and does not actually do any training. Inside the trainable function, when you do `trainer.fit` with a Trainer that contains the...

Rank Zero Deprecation

Hey @lcaquot94, thanks for reporting this! Indeed, we should change this- a PR would be much appreciated!