[EM] Enable access to the number of batches. #10691
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
trivialfis
force-pushed
the
ext-n-batches
branch
from
9e99eb6 to
745ce87
Compare
|
Okay, so the CUDA error is observable on the CI as well. |
|
Note to myself: The GPU test is invalid in the master branch, update prediction cache simply returns false due to external memory not being supported. |
|
|
- Expose `NumBatches` in `DMatrix`. - Small cleanup for removing legacy CUDA stream and force CUDA context initialization.
trivialfis
force-pushed
the
ext-n-batches
branch
from
0c6923f to
ae85e0d
Compare
|
cc @rongou . |
rongou
approved these changes
Aug 16, 2024
| // action might cause an insufficient driver version error for some reason. Lastly, it | ||
| // should work with CUDA 12.5 without any action being taken. | ||
|
|
||
| // dh::DefaultStream().Sync(); |
There was a problem hiding this comment.
Yes, it's needed for 12.2, but it seems fine with 12.5. However, currently the CI is running 11.8, which doesn't work with the initialization.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
NumBatchesinDMatrix.force CUDA context initialization.The CUDA uninitialized error was observed in CTK 12.2, tested CTK 12.5 on a different machine but it's not observable there. Let's keep the stream sync here to avoid potential errors, it only runs at the beginning of a thread.
A subtle difference is that, after the PR, if an external DMatrix contains only a single batch, then the
SingleColBlockwould return true. Currently, it returns true regardless of where it actually contains a single block when the data type is external memory.update: Removed the force initialization for now due to CI running CTK 11.8.