WSL2环境下由于找不到CUDA_CUDNN_LIBRARY无法通过pip安装mirage

[WuTianyi321]

在micromamba环境下使用pip install mirage一直报

CMake Error: The following variables are used in this project, but they are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake files:
CUDA_CUDNN_LIBRARY

的错误，实际上echo $CUDA_CUDNN_LIBRARY结果为'/envs/torch/lib/libcudnn.so'。

wsl2系统为Debian GNU/Linux 12
nvcc -V命令结果为

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Tue_Feb_27_16:19:38_PST_2024
Cuda compilation tools, release 12.4, V12.4.99
Build cuda_12.4.r12.4/compiler.33961263_0

[SpiritedAwayCN]

Currently, the mirage installation has not been fully tested on WSL.

Could you try installing mirage from source, and show your full cmake logs?
In my Ubuntu Linux docker, cmake will find CUDA and output following logs:

-- CMake Version: 3.30.2
-- CUDART: /usr/local/cuda/lib64/libcudart.so
-- CUDA Driver: /usr/local/cuda/lib64/stubs/libcuda.so
-- NVRTC: /usr/local/cuda/lib64/libnvrtc.so
-- CUDA_INCLUDE_DIRS=/usr/local/cuda/include
-- Found CUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda
-- Found CUDA_CUDA_LIBRARY=/usr/local/cuda/lib64/stubs/libcuda.so
-- Found CUDA_CUDART_LIBRARY=/usr/local/cuda/lib64/libcudart.so
-- Found CUDA_NVRTC_LIBRARY=
-- Found CUDA_CUDNN_LIBRARY=/usr/lib/x86_64-linux-gnu/libcudnn.so
-- Found CUDA_CUBLAS_LIBRARY=/usr/local/cuda/lib64/libcublas.so
-- CUDA_INCLUDE_DIR=/usr/local/cuda/include
-- Z3_FOUND: 1
-- Found Z3 4.13.1.0
-- Z3_DIR: /path/to/mirage/deps/z3/build
-- Using the multi-header code from /path/to/mirage/deps/json/include/
-- Configuring done (0.2s)
-- Generating done (0.2s)
-- Build files have been written to: /path/to/mirage/build

According to the CMAKE scripts:
https://github.com/mirage-project/mirage/blob/main/cmake/cuda.cmake#L68

CUDA_CUDNN_LIBRARY can be set when finding CUDA Toolkit. You can also modify this cmake file to check whether CUDA_TOOLKIT_ROOT_DIR is properly set. Thanks for your feedback!