Fail to install CUDA, Flux, CuArrays etc, pls advise - Julia packages

776 views Asked by At

I am new to Julia, when trying to add CUDA, comes the error as following. It seems meaning the CuArrays is not in right condition, so I add the CuArrays from the github, then it comes another error about Adapt. When adding Adapt, it seems fine, but when rolling back, the Adapt issue comes again. Just like rabbit holes in Alice's dream, I am stucked here and don't know what to do to make it work. Please advise, Thanks,

BTW, the CUDA path is set.

OS:Win10
JuliaPro 1.5.0
GPU: NVIDIA MX150

CUDA.jl

(@JuliaPro_v1.5.0-1) pkg> add https://github.com/JuliaGPU/CUDA.jl
   Updating git-repo `https://github.com/JuliaGPU/CUDA.jl`
  Resolving package versions...
ERROR: Unsatisfiable requirements detected for package CUDAnative [be33ccc6]:
 CUDAnative [be33ccc6] log:
 ├─possible versions are: [0.7.0, 0.8.0-0.8.10, 0.9.0-0.9.1, 0.10.0-0.10.1, 1.0.0-1.0.1, 2.0.0-2.0.1, 2.1.0-2.1.3, 2.2.0-2.2.1, 2.3.0-2.3.1, 2.4.0, 2.5.0-2.5.5, 2.6.0, 2.7.0, 2.8.0-2.8.1, 2.9.0-2.9.1, 2.10.0-2.10.2, 3.0.0-3.0.4, 3.1.0, 3.2.0] or uninstalled
 ├─restricted to versions * by an explicit requirement, leaving only versions [0.7.0, 0.8.0-0.8.10, 0.9.0-0.9.1, 0.10.0-0.10.1, 1.0.0-1.0.1, 2.0.0-2.0.1, 2.1.0-2.1.3, 2.2.0-2.2.1, 2.3.0-2.3.1, 2.4.0, 2.5.0-2.5.5, 2.6.0, 2.7.0, 2.8.0-2.8.1, 2.9.0-2.9.1, 2.10.0-2.10.2, 3.0.0-3.0.4, 3.1.0, 3.2.0]
 └─restricted by compatibility requirements with LLVM [929cbde3] to versions: uninstalled — no versions left
   └─LLVM [929cbde3] log:
     ├─possible versions are: [0.9.0-0.9.15, 1.0.0, 1.1.0-1.1.1, 1.2.0, 1.3.0-1.3.4, 1.4.0-1.4.1, 1.5.0-1.5.2, 1.6.0, 1.7.0, 2.0.0, 3.0.0] or uninstalled
     └─restricted to versions 3 by CUDA [052768ef], leaving only versions 3.0.0
       └─CUDA [052768ef] log:
         ├─possible versions are: 1.3.0 or uninstalled
         └─CUDA [052768ef] is fixed to version 1.3.0

CUDAnative

(@JuliaPro_v1.5.0-1) pkg> add https://github.com/JuliaGPU/CUDAnative.jl
   Updating git-repo `https://github.com/JuliaGPU/CUDAnative.jl`
  Resolving package versions...
ERROR: Unsatisfiable requirements detected for package Adapt [79e6a3ab]:
 Adapt [79e6a3ab] log:
 ├─possible versions are: 2.3.0 or uninstalled
 ├─restricted to versions [0.4, 1] by CUDAnative [be33ccc6] — no versions left
 │ └─CUDAnative [be33ccc6] log:
 │   ├─possible versions are: 3.1.0 or uninstalled
 │   └─CUDAnative [be33ccc6] is fixed to version 3.1.0
 └─Adapt [79e6a3ab] is fixed to version 2.3.0

Adapt

(@JuliaPro_v1.5.0-1) pkg> add https://github.com/JuliaGPU/Adapt.jl
   Updating git-repo `https://github.com/JuliaGPU/Adapt.jl`
  Resolving package versions...
No Changes to `C:\Users\eric1\.julia\environments\JuliaPro_v1.5.0-1\Project.toml`
No Changes to `C:\Users\eric1\.julia\environments\JuliaPro_v1.5.0-1\Manifest.toml`

CUDA Path GPU running

1

There are 1 answers

3
Przemyslaw Szufel On BEST ANSWER

You should be installing packages:

(@JuliaPro_v1.5.0-1) pkg> add CUDA

If you are doing instead add https://github.com/JuliaGPU/CUDA.jl you are installing the latest development version (master branch) from the repository. Since it is a development (current-state-of-source-code) version quite likely something might not work.

Now, I recommend you to remove all those dev package versions that you have installed (so run rm package manager commad) and then reinstall their release versions rather than dev versions.

There is a good change that after that you will not have a dependency clash.