search

FluxML / Torch.jl

Sensible extensions for exposing torch in Julia.
Other
213 stars 15 forks source link

Failed installing Torch.jl #50

Open JinraeKim opened 2 years ago

JinraeKim commented 2 years ago

Simply add Torch results in this error:

ERROR: LoadError: InitError: could not load library "/home/jinrae/.julia/artifacts/d6ce2ca09ab00964151aaeae71179deb8f9800d1/lib/libdoeye_caml.so"
libcufft.so.10: cannot open shared object file: No such file or directory

with nvidia-smi output said: Driver Version: 460.106.00 CUDA Version: 11.2 GPU: GeForce RTX 2070

overshiki commented 2 years ago

+1 I also encountered this problem, is there anyone who has a clue of what's going on?

DhairyaLGandhi commented 2 years ago

Can you try adding master? We can cut a release to reinstate the functionality.

On Sun, Oct 30, 2022, 19:31 overshiki @.***> wrote:

+1 I also encountered this problem, is there anyone who has a clue of what's going on?

— Reply to this email directly, view it on GitHub https://github.com/FluxML/Torch.jl/issues/50#issuecomment-1296267284, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJOZVVJ32AHL3BEZQYPMAK3WFZ5VJANCNFSM5JA5I2PA . You are receiving this because you are subscribed to this thread.Message ID: @.***>

overshiki commented 2 years ago

Hi @DhairyaLGandhi Do you mean this:

add https://github.com/FluxML/Torch.jl.git

I tried this, but unfortunately, it does not work either. It raises the same error as add Torch. Namely, the libdoeye_caml.so file is missing.

To be more specific, the error I faced is slightly different from that of @JinraeKim :

ERROR: LoadError: InitError: could not load library "/home/le/.julia/artifacts/d6ce2ca09ab00964151aaeae71179deb8f9800d1/lib/libdoeye_caml.so"
libcuda.so.1: cannot open shared object file: No such file or directory

And I'm on a platform without a GPU(I just want to check the CPU version)

Thanks a lot if this problem could be solved, I'm doing some experiments to use PyTorch models in a Julia environment, and currently I'm highly relying on PyCall .

ToucheSir commented 2 years ago

I don't think things will work until all the points in https://github.com/FluxML/Torch.jl/issues/54#issuecomment-1264575111 are complete, unfortunately. In the meantime, check out https://github.com/rejuvyesh/PyCallChainRules.jl and possibly https://github.com/lorenzoh/PyNNTraining.jl if you want better integration with PyTorch models.

overshiki commented 2 years ago

Cool, thanks! I'll check out these two packages. And also looking forward to the completion of Torch.jl!