====== Installed Tools and Libraries ======


===== CUDA and CuDNN ====

On a gpuhost, the drivers are installed (check it with nvidia-smi), but not the CUDA toolkit. 

Versions 9.0 and 10.0 are available at:<code>/scratch/gpuhost6/pguhur/bin/cuda-xx</code>

Versions 8.0 and 10.1 are available at: <code>/scratch/curan/rgarciap/bin/cuda-xx</code>

Versions 10.0, 10.2, 11.0 and 11.2 are available at: <code>/scratch/clear/thoth/cuda/cuda-xx</code>

You can easily export them in your environment by adding the following lines into your .bashrc (or .zshrc):

<code>
if [[ "`hostname`" == gpuhost* ]]; then
  export CUDA_ROOT=/scratch/clear/thoth/cuda/cuda-11.0
  export LD_LIBRARY_PATH=$CUDA_ROOT/lib64
  export CUDA_HOME=$CUDA_ROOT
  export LIBRARY_PATH=$LD_LIBRARY_PATH
  export PATH=$CUDA_ROOT/bin:$PATH
fi
</code>

Contributed by **Pierre-Louis Guhur**, 06/06/2019


===== Datasets =====

Downloading and uncompressing large datasets take a serious amount of time. Here are some already installed datasets:

  * [[https://cs.stanford.edu/~kaichun/partnet/|PartNet]]: is available at:
<code>/scratch/gpuhost6/pguhur/datasets/partnet/</code>

  * [[http://www.image-net.org/challenges/LSVRC/2012/|ImageNet ILSVRC2012]]: is available at:
<code>/scratch/curan/hleterme/datasets/ImageNet2012</code>

Contributed by **Hubert Leterme**, 06/02/2020, revised 2022-09-17

===== py-faster-rcnn =====

If you are struggling to compile py-faster-rcnn on ubuntu, try more up-to-date version:
https://github.com/Austriker/py-faster-rcnn
It should compile (modulo probably matcaffe that I didn't tested).

Contributed by **Konstantin Shmelkov**, 22/09/2016