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
Fuse split evaluation kernels #8026
Conversation
Depth 8
Depth 20
Depth 8 lossguide
|
One of the dask early stopping tests started to fail but this is only due to slight changes in split evaluation. The validation score continues to decrease beyond maximum number of iterations. This PR has resulted in very small changes to gbm-bench training loss results, some better, some worse. I attribute this to floating point ordering. In particular, changing the block size for split evaluation from 256->32 has an impact. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good to me!
Benchmarking to follow.