Skip to content

Latest commit

 

History

History
75 lines (67 loc) · 1.63 KB

INSTALL.md

File metadata and controls

75 lines (67 loc) · 1.63 KB
Raw

INSTALLATION OF NN PROJECT

Installing Anaconda

After downloading the Anaconda installer, run the following command from a terminal:

$ bash Anaconda-2.x.x-Linux-x86[_64].sh

You then need to edit your .bashrc file to include the ECHO $PATH of where you installed Anaconda. You normally get notified about this at the end of the instalation.

Using Conda

  • conda create --name nn python=3
  • source activate nn
  • (opt) pip install ipython
  • pip install scipy
  • pip install numpy
  • pip install pandas
  • conda install xarray dask netCDF4 bottleneck
  • pip install matplotlib
  • pip install --upgrade https://github.com/Theano/Theano/archive/master.zip
  • pip install keras
  • pip install sklearn
  • pip install h5py
  • conda install mkl-service
  • conda install netCDF4
  • for visualization :
    • pip install pydot-ng
    • conda install graphviz
  • Modify your $HOME/.theanorc file:
[global]
floatX = float32
device = cpu

[nvcc]
fastmath = True
  • Modify your ~/.keras/keras.json file:
{
    "floatx": "float32",
    "epsilon": 1e-07,
    "backend": "theano",
    "image_dim_ordering": "th"
}
  • Test if it works : ./script_train_test.py

Using GPU

  • Modify your $HOME/.theanorc file:
[global]
floatX = float32
device = gpu

[nvcc]
fastmath = True

[lib]
cnmem = 1

[dnn]
enabled = True
  • Modify your $HOME/.bashrcfile to add the path and library path of cuda : e.g
export CPATH="/usr/local/cuda/include:$CPATH"
export PATH="/usr/local/cuda/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda/lib64:$LD_LIBRARY_PATH"