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
[jvm-package] scala rabit tracker underflow #1925
Comments
Here is a working example documenting the problem https://gist.github.com/geoHeil/5e355bc7666294574ae9432dd912806f @xydrolase hope this helps to track down the problem |
Will look into this. Thanks. |
Thanks for providing the example for reproducing the error. Did you consistently run into the same buffer issue? I can kind of guess where and why it threw that Exception, because the parser for worker command is not exhaustive and may fail in some conditions. I'll use your stacktrace to see if I can pinpoint the issue, while continue playing with your example to see if I can replicate it or not. |
This would happen if the worker sends a "print" command message in multiple packets. The "print" command is special as it contains a string, which is the message to print of course. The validator for commands only verifies the common part, but ignores the optional message. And because the validation is generic to all commands, it does not include any additional logic to the "print" command. I'll fix this this week. |
@xydrolase is there any activity with this issue? |
Sorry, pretty hectic recently. It has definitely been 2 weeks since my original promise of "this week". It's an easy fix, so I will for sure allocate some time Friday to submit a PR to fix this issue. |
@geoHeil I've fixed it locally. Will push it tomorrow after merging upstream commits. |
@geoHeil I've submitted the PR that hopefully fixes the issue for good. |
@xydrolase thanks a lot. |
scala rabit tracker is crashing with an underflow
Environment info
Operating System: osx 10.12.2
Compiler: gcc6
Package used (python/R/jvm/C++): jvm-package
xgboost
version used: current master branchIf installing from source, please provide
git rev-parse HEAD
)The good thing is the scala implementation seems to be a lot quicker - donwside is: it is crashing.
edit
I noticed that the described problem only occurs when using XGBoostEstimator in a cross validation pipe. If accessed directly (still in a pipeline without cross validation) it seems to work just fine
and
simplPipe.fit(train).transform(test).show
works just fineThe text was updated successfully, but these errors were encountered: