search

lab-medvedeva / GADES-main

GNU General Public License v3.0
0 stars 1 forks source link

Error in GPU Mode for Kendall Distance Calculation Returns Zero Matrix #8

Open TorikoP opened 1 week ago

TorikoP commented 1 week ago

Hi! I've followed the local installation instructions for the GADES package:

git clone https://github.com/lab-medvedeva/GADES-main.git cd GADES-main Rscript install.R

When executing the install.R script, I encountered the following error:

Error in contrib.url(repos, type) : trying to use CRAN without setting a mirror Calls: install.packages -> startsWith -> contrib.url Execution halted

To resolve the issue, I added the following line to the install.R script: options(repos = structure(c(CRAN = "https://cran.r-project.org")))

After successfully installing the package, I tried to compute a Kendall distance matrix using the function: mtrx_distance(matrix, metric = 'kendall', type='gpu', sparse = F, write=T)

However, the function returns a matrix containing only zeros

When I run the same function in cpu mode, it works as expected and returns a proper Kendall distance matrix:

mtrx_distance(matrix, metric = 'kendall', type='cpu', sparse = F, write=T)

It seems that there may be an issue with GPU recognition or GPU computation. I would appreciate guidance on identifying what might be going wrong with the GPU mode

akhtyamovpavel commented 1 week ago

Hi, @TorikoP !

Thank you for providing the issue!

Could you provide the additional details of your environment? We are interested in the following details:

  1. Operating System (Linux, Windows, MacOS)
  2. Development Environment (Docker or Bare Metal)
  3. CUDA version installed in your environment (If the answer for the question 2 is "Bare Metal")
  4. Your GPU.

The answer to the questions above will help us the issue as soon as possible!

TorikoP commented 1 week ago

Here is all details:

  1. Linux
  2. Bare Metal
  3. The CUDA version is 12.6
  4. GPUs: Quadro RTX 5000 (2 units) and NVIDIA GeForce GTX TITAN X