Skip to content
This repository has been archived by the owner on Nov 4, 2024. It is now read-only.

MLDataDevices requires cuDNN to load cuda functionality #79

Closed
vpuri3 opened this issue Oct 3, 2024 · 2 comments
Closed

MLDataDevices requires cuDNN to load cuda functionality #79

vpuri3 opened this issue Oct 3, 2024 · 2 comments

Comments

@vpuri3
Copy link
Contributor

vpuri3 commented Oct 3, 2024 

julia> using CUDA, MLDataDevices

julia> MLDataDevices.loaded(MLDataDevices.CUDADevice)
false

julia> using LuxCUDA

julia> MLDataDevices.loaded(MLDataDevices.CUDADevice)
true
julia> using CUDA, MLDataDevices

julia> using cuDNN

julia> MLDataDevices.loaded(MLDataDevices.CUDADevice)
true
@avik-pal
Copy link
Member

avik-pal commented Oct 3, 2024

Yes this is an intentional decision. We anyways can't do any "ML" on cuda gpus without that being loaded.

@avik-pal avik-pal closed this as not planned Won't fix, can't repro, duplicate, stale Oct 3, 2024
@vpuri3
Copy link
Contributor Author

vpuri3 commented Oct 3, 2024

The docs should then be modified to say that the trigger package is LuxCUDA or cuDNN and not CUDA.jl

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@avik-pal @vpuri3