使用Conda安装RAPIDS 23.08失败，提示PackagesNotFoundError

问题概述 #

执行RAPIDS官方安装命令时，环境求解失败，提示找不到rapids=23.08包。

使用的安装命令 #

conda create --solver=libmamba -n rapids-23.08 -c rapidsai -c conda-forge -c nvidia rapids=23.08 python=3.10 cuda-version=12.0

报错信息 #

Channels:
 - rapidsai
 - conda-forge
 - nvidia
 - defaults
Platform: win-64
Collecting package metadata (repodata.json): - DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): conda.anaconda.org:443
DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): repo.anaconda.com:443
DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): repo.anaconda.com:443
DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): repo.anaconda.com:443
DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): repo.anaconda.com:443
DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): conda.anaconda.org:443
\ DEBUG:urllib3.connectionpool:https://repo.anaconda.com:443  "GET /pkgs/main/noarch/repodata.json HTTP/1.1" 304 0
DEBUG:urllib3.connectionpool:https://repo.anaconda.com:443  "GET /pkgs/r/win-64/repodata.json HTTP/1.1" 304 0
DEBUG:urllib3.connectionpool:https://repo.anaconda.com:443  "GET /pkgs/r/noarch/repodata.json HTTP/1.1" 304 0
DEBUG:urllib3.connectionpool:https://repo.anaconda.com:443  "GET /pkgs/main/win-64/repodata.json HTTP/1.1" 304 0
/ DEBUG:urllib3.connectionpool:https://conda.anaconda.org:443  "GET /rapidsai/noarch/repodata.json HTTP/1.1" 304 0
- DEBUG:urllib3.connectionpool:https://repo.anaconda.com:443  "GET /pkgs/msys2/win-64/repodata.json HTTP/1.1" 304 0
DEBUG:urllib3.connectionpool:https://repo.anaconda.com:443  "GET /pkgs/msys2/noarch/repodata.json HTTP/1.1" 304 0
\ DEBUG:urllib3.connectionpool:https://conda.anaconda.org:443  "GET /rapidsai/win-64/repodata.json HTTP/1.1" 304 0
done
Solving environment: failed

PackagesNotFoundError: The following packages are not available from current channels:

  - rapids=23.08*

Current channels:

  - https://conda.anaconda.org/rapidsai 
  - https://conda.anaconda.org/conda-forge 
  - https://conda.anaconda.org/nvidia 
  - defaults

To search for alternate channels that may provide the conda package you're
looking for, navigate to

    https://anaconda.org 

and use the search bar at the top of the page.

系统环境 #

• OS：Windows Server 2022 Datacenter
• Python：3.10.12
• conda：23.7.3
• CUDA：12.0

解决方案 #

1. 添加Windows专属频道 #

RAPIDS的Windows版本包托管在rapidsai-windows频道，修改安装命令添加该频道：

conda create --solver=libmamba -n rapids-23.08 -c rapidsai-windows -c rapidsai -c conda-forge -c nvidia rapids=23.08 python=3.10 cuda-version=12.0

2. 安装核心组件替代完整元包 #

如果仍找不到rapids元包，可直接安装核心组件（如cuDF、cuML）：

conda create --solver=libmamba -n rapids-23.08 -c rapidsai-windows -c conda-forge -c nvidia cudf=23.08 cuml=23.08 python=3.10 cuda-version=12.0

3. 配置频道优先级 #

确保相关频道优先级高于defaults，避免包版本冲突：

conda config --add channels rapidsai-windows
conda config --add channels rapidsai
conda config --add channels conda-forge
conda config --add channels nvidia
conda config --set channel_priority strict

配置完成后重新执行安装命令。

内容的提问来源于stack exchange，提问作者Tio