You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
I am trying to run Flux on a server (digital research alliance of canada). It is on notebook through Jupyter Hub. I use the CUDA.set_runtime_version!("local") to access the local cuda runtime, and restart Julia. Then Cuda works (can add number to CuArray), but pre-compiling cuDNN fails. Therefore using Flux fails with error ERROR: LoadError: ArgumentError: invalid version string: local. using Lux fails with the same error.
Stacktrace:
I have yet to hear back from the system admin for this, and also these versions are the most recent ones I have access to on the server. Also let me know if this issue belongs somewhere else (i.e. it is not an issue but something I am doing wrong). Thanks!
To reproduce
The Minimal Working Example (MWE) for this bug:
using Pkg
Pkg.activate("./")
Pkg.add("CUDA")
Pkg.add("cuDNN")
using CUDA
CUDA.set_runtime_version!("local")
Describe the bug
I am trying to run Flux on a server (digital research alliance of canada). It is on notebook through Jupyter Hub. I use the
CUDA.set_runtime_version!("local")
to access the local cuda runtime, and restart Julia. Then Cuda works (can add number to CuArray), but pre-compiling cuDNN fails. Thereforeusing Flux
fails with errorERROR: LoadError: ArgumentError: invalid version string: local
.using Lux
fails with the same error.Stacktrace:
I have yet to hear back from the system admin for this, and also these versions are the most recent ones I have access to on the server. Also let me know if this issue belongs somewhere else (i.e. it is not an issue but something I am doing wrong). Thanks!
To reproduce
The Minimal Working Example (MWE) for this bug:
-> restart Julia
Manifest.toml
Expected behavior
Load cuDNN (really,
using Lux
, orusing Flux
to not fail)Version info
Details on Julia:
Julia 1.8.1
IJulia kernel 1.8
CUDA.jl 4.0.1
CUDA runtime 11.7, local installation
CUDA driver 11.8
Flux v0.13.13 (or Lux v0.4.40)
Details on CUDA:
The text was updated successfully, but these errors were encountered: