-
Due to changes in RMM, there are a few notebooks that need to be updated to the new API. Alerting those who maintain the respective notebooks.
Changes Readme: https://github.com/rapidsai/rmm#han…
-
We should use the new [`rmm::exec_policy`](https://github.com/rapidsai/rmm/blob/2f85111679bf50ccb68fe36d935cbf452e427d84/include/rmm/exec_policy.hpp) instead of the default `thrust::cuda::par` policy,…
-
Hello,
I use RUCKZUCK a lot and it is really great! Thank you very much! Coupled with my RMM, it allows me to launch automation for searching and installing updates.
I regularly encounter proble…
-
**Describe the bug**
I am observing we use `426 Mib` memory outside the pool when training/using a cuML model.
See MRE below (trace [here](https://gist.github.com/VibhuJawa/cb93401d6a6c1ed7…
-
PR #946 introduces a workaround for an issue in cuda-python that fetches a hardcoded CUDA runtime version. https://github.com/NVIDIA/cuda-python/issues/16
Once that upstream issue is resolved, we s…
bdice updated
6 months ago
-
Scalars can be null and contain a validity bool. The constructors for the various scalars take an `is_valid` parameter for this purpose. But the factory functions themselves don't let you specify …
-
Helps with finding why a client or site is yellow/red
Any task with info should not affect green status
-
**Is your feature request related to a problem? Please describe.**
want to use cuda multi stream (pool) read a big file.
**Describe the solution you'd like**
just add a new `multibyte_split` met…
-
Working on https://github.com/rapidsai/raft/pull/1748, I realized the need to introduce types that allow for memory access from host and device both, as more algorithms need to support allocating eith…
-
**Is your feature request related to a problem? Please describe.**
Currently when librmm is built, it accepts a CMake option `CUDA_STATIC_RUNTIME` to determine whether to link to cudart statically or…
vyasr updated
11 months ago