pluskid / Mocha.jl

Deep Learning framework for Julia
Other
1.29k stars 254 forks source link

gpubackend() error #164

Closed vetrianka closed 8 years ago

vetrianka commented 8 years ago

Hello,

I guess it is a silly question, but when I am trying to initialize GPU backend, it raises an error:

julia> ENV["MOCHA_USE_CUDA"] = "true" "true"

julia> using Mocha WARNING: Method definition info(Any...) in module Base at util.jl:334 overwritte n in module Logging at C:\Users\ALEXEY.julia\v0.4\Logging\src\Logging.jl:61. WARNING: Method definition warn(Any...) in module Base at util.jl:364 overwritte n in module Logging at C:\Users\ALEXEY.julia\v0.4\Logging\src\Logging.jl:61. Configuring Mocha...

julia> backend = GPUBackend() Mocha.GPUBackend

julia> init(backend) ERROR: error compiling init: error compiling create: could not load library "" The specified module could not be found.

Would you please help me? Here are version numbers: Windows 8.1 64 Julia 0.4.0 Cuda 7.5.18 Cudnn 3.0 Kernels compiled with Visual Studio 12

greenflash1357 commented 8 years ago

I guess your cuBLAS library is not found.

Compare /src/cuda/cublas.jl line 39: const libcublas = Libdl.find_library(["cublas64_70.dll", "cublas64_65.dll", cublas32_70.dll", "cublas32_65.dll"], [""])

Your CUDA version is 7.5.

You can add "cublas64_75.dll" to the search list, but I'm not sure wether CUDA 7.5 is really supported by Mocha.

pluskid commented 8 years ago

@greenflash1357 Thanks! @vetrianka Yes Julia Libdl basically eat all the error messages when loading dynamic libraries, making it very difficult to debug. Can you try @greenflash1357's suggestion and see if it works? Please let us know so that we could add the fix to the current code repo. Thanks!

vetrianka commented 8 years ago

Thank you very much for your replies. It seems the issue is resolved.