IsolationKernel / Codes

Other
50 stars 15 forks source link

Segmentation fault running psKC #2

Open tmadl opened 2 years ago

tmadl commented 2 years ago

Hi, thanks for open sourcing this!

I get a Segmentation fault trying to run psKC (both release and debug) and cannot figure out where the issue is. Data: make_circles from sklearn. Tried different growth rates and thresholds and set and sample sizes; always end up with:

`Preprocessing data ... gpu::compute_distance:

Segmentation fault `

Could you please share the exact setup under which you compiled and made it work? I.e., gcc version, cuda version, any other libraries & version that might have caused the error?

Thanks in advance - would love to try this out & cite it if it works well

PS: a conda installable Python version of psKC would be hugely helpful!!

jrwells-research commented 1 year ago

Sorry for the delay in replying.

Here is the additional requested infomation.

Development Environment

These source code releases are developed on an Arch Linux system. The GPU codes are tested on a Nvidia GeForce 1080 Ti. While some codes have been added to support Microsoft Windows, it is not a targetted platform.

The following tools are used during the develepment:

Note: You do not need any of the QT libraries (e.g., GUI or CLI). To install qmake, on the Arch Linux, install qt5-base package.

Also note that if you do use the CUDA library, you will need to install the correct GCC version for CUDA.

To build the system use the following commands:

  1. Go to the subdirectory containing the '.qmake.conf' file.
  2. Create a sub-directory called 'build'.
  3. Go to the 'build' sub-directory.
  4. Execute 'qmake ..'
  5. Execute 'make'

And the individual executeables are found in the corresponding sub-directories of the source code within the build sub-directory. Note that both 'debug' and 'release' version are built.

Temp Directory

Warning: Make sure that you have plenty of disk space for the storage of temp files. Depending on what parameters being used, the disk usage could be up and over 1 TiB.