RAPIDS Docs

This site uses Just the Docs, a documentation theme for Jekyll.

Search

RSN 20 - Please Pin cuda-python to 11.7.0

Author RAPIDS Ops
Status

Completed
Topic Platform Support Change
RAPIDS Version v22.02, v22.04, and v22.06
Created 15 July 2022
Updated 05 September 2023

Overview

We have recently become aware that cuda-python 11.7.1 has incorporated changes that makes it incompatible with current and legacy versions of RAPIDS when you try to install RAPIDS. You will need to pin cuda-python to 11.7.0 in conda during installation, and potentially after. An affected installation with the 11.7.1 issue results in this error being thrown:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/opt/conda/lib/python3.9/site-packages/cudf/__init__.py", line 5, in <module>
    validate_setup()
  File "/opt/conda/lib/python3.9/site-packages/cudf/utils/gpu_utils.py", line 20, in validate_setup
    from rmm._cuda.gpu import (
  File "/opt/conda/lib/python3.9/site-packages/rmm/__init__.py", line 16, in <module>
    from rmm import mr
  File "/opt/conda/lib/python3.9/site-packages/rmm/mr.py", line 14, in <module>
    from rmm._lib.memory_resource import (
  File "/opt/conda/lib/python3.9/site-packages/rmm/_lib/__init__.py", line 15, in <module>
    from .device_buffer import DeviceBuffer
  File "device_buffer.pyx", line 1, in init rmm._lib.device_buffer
TypeError: C function cuda.ccudart.cudaStreamSynchronize has wrong signature (expected __pyx_t_4cuda_7ccudart_cudaError_t (__pyx_t_4cuda_7ccudart_cudaStream_t), got cudaError_t (cudaStream_t)

Status

Currently, we have these two resolution paths to offer:

Resolution Path 1: You can upgrade your RAPIDS version to `v22.06.01`.
- We've made it easy to upgrade: simply use our Release selector and install `v22.06` or rerun you `v22.06` installation script. It will automatically grab `v22.06.01` via Conda or Docker.

Resolution Path 2: If you can NOT upgrade, pin `cuda-python` to `11.7.0` in RAPIDS `v22.02`, `v22.04`, and `v22.06`,
- You can pin `cuda-python` to `11.7.0` by using `cuda-python=11.7.0` when doing an install.
- Example command: `conda install -c nvidia cuda-python=11.7.0`.

Older versions of RAPIDS may also be affected and users should attempt to upgrade or pin.

Impact

This issue impacts users installing conda packages of RAPIDS 22.02 through 22.06, which depend on cuda-python. This issue also affects existing conda installations of RAPIDS versions 22.02 - 22.06 if the user installs additional packages depending on cuda-python that result in an update to cuda-python>11.7.0.