Additionally I haven’t been able to get CuArrays to work properly, and I have yet to find a Flux tutorial that I can run succesfully in its entirety. I am running my code in a conda environment on a server with a GPU without root access, and I’ve been able to run deep learning models using Pytorch and fastai without any problems in the same conda environment.
The server is running on Ubuntu 16.04.6 LTS (GNU/Linux 4.15.0-58-generic x86_64).
This is information I get on the GPU when I run lshw -C display from the terminal:
(GPU Information: Click to expand)
description: VGA compatible controller
product: GV102
vendor: NVIDIA Corporation
physical id: 0
bus info: pci@0000:01:00.0
version: a1
width: 64 bits
clock: 33MHz
capabilities: vga_controller bus_master cap_list rom
configuration: driver=nvidia latency=0
resources: irq:126 memory:a3000000-a3ffffff memory:90000000-9fffffff memory:a0000000-a1ffffff ioport:3000(size=128)memory:c0000-dffff
WARNING: output may be incomplete or inaccurate, you should run this program as super-user.
I am running Julia Version 1.0.3 (conda-forge-julia release). These is the output from running using Pkg; Pkg.installed():
(Pkg info: Click to expand)
Dict{String,Union{Nothing, VersionNumber}} with 12 entries:
"CSV" => v"0.5.12"
"Missings" => v"0.4.2"
"Metalhead" => v"0.3.0"
"DataFramesMeta" => v"0.5.0"
"StatsBase" => v"0.32.0"
"CuArrays" => v"1.2.1"
"IJulia" => v"1.20.0"
"Flux" => v"0.9.0"
"CUDAnative" => v"2.3.1"
"DataFrames" => v"0.19.4"
"FilePaths" => v"0.8.0"
"Pandas" => v"1.3.0"
I don’t have a ton of experience with Julia, but I am really interested in getting setup using Julia for some of my deep learning projects. As the above error messages are unintuitive to me, any help would be very much appreciated.
(Note: I split this message up into parts due to the character limit on discourse)