Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[EM] Treat ellpack as dense matrix when there's no compression. #10870

Merged
merged 6 commits into from
Oct 3, 2024

Conversation

trivialfis
Copy link
Member

@trivialfis trivialfis commented Oct 1, 2024

This enables XGBoost to use shared memory to build histograms. In addition, it lowers memory usage and speeds up external memory for data that cannot be compressed using the ellpack format.

  • From now on, only datasets with at least one missing value for every sample are considered sparse.
  • We need to distinguish fully dense, mostly dense, and sparse. This is similar to what the CPU implementation currently does.
  • Some cleanups

Compression info.

Row stride.

Add test.

cleanup.

Add tests.

name.

Cleanup.

comment.

used n_symbols.

cleanup.

Histogram.

Split up tests.

Write compressed.

Comment.

from ghist.

Work on ellpack -> ghist.

look into adapter case.

from data.

cleanup.

start working on tests.

check.

From adapter.

to ghist.

dense.

empty.

feature groups.

feature group.

histogram.

rename.

CPU.

cleanup.

Enable get cut.

Fix test.
cleanup.

Fix.

test.

fix.
@trivialfis
Copy link
Member Author

trivialfis commented Oct 3, 2024

cc @rongou

It's a big PR, I'm happy to do online review if needed.

@trivialfis trivialfis merged commit 1b4c5fb into dmlc:master Oct 3, 2024
30 checks passed
@trivialfis trivialfis deleted the ext-ellpack-as-dense branch October 3, 2024 16:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants