Adapt to rmm logger changes#499
Conversation
9e198e4 to
9e151ff
Compare
|
The arm wheel test is running into https://bugzilla.redhat.com/show_bug.cgi?id=1722181. A good reference discussion is in pytorch/pytorch#2575. dmlc/xgboost#8488 is an even better example of this issue, the torch one is closely related but slightly different. Ultimately the problem is that thread-local storage is being overallocated by some combination of the libraries that we are using before OpenMP is loaded. We may be able to fix this by changing the import order of packages, but we'll need an arm machine to test that since this test is only failing on SBSA (I assume architectural differences lead to different TLS availability). It looks like we have precedent for using LD_PRELOAD in cuml, so I'm going to toss that into cuvs tests as well. Unfortunately because it's a wheel test we have to load the library that is found inside wheels, so I just added an import in conftest. |
|
/merge |
This PR adapts to breaking changes in rmm in rapidsai/rmm#1722.