The following contains instructions for a 'full' installation of nanonet including currennt, as required for training networks. If you wish to use nanonet only for basecalling please refer to the file README.md.
For convenience Oxford Nanopore Technologies provides binary packages for Ubuntu. These will perform a full install of all components including currennt.
wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_7.5-18_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1404_7.5-18_amd64.deb
sudo apt-get update
sudo apt-get -y install cuda-cudart-7-5 cuda-cublas-7-5 netcdf-bin libboost-all-dev
wget https://github.com/nanoporetech/currennt/releases/download/v0.2-rc1-2/python-netcdf4_1.2.3-1_amd64.deb \
https://github.com/nanoporetech/currennt/releases/download/v0.2-rc1-2/ont-currennt-0.2.1-3-trusty.deb \
https://github.com/nanoporetech/nanonet/releases/download/v1.1.2/python-nanonet_1.1.3-1_amd64.deb
sudo dpkg -i python-netcdf4_1.2.3-1_amd64.deb ont-currennt_0.2.1-2-trusty_amd64.deb python-nanonet_1.1.3-1_amd64.deb
# expect an error here about missing prerequisite packages, which can be corrected with:
sudo apt-get -f install
The following was tested to work on a clean Ubuntu 14.04 Vagrant virtual machine.
Installation of currennt
Oxford Nanopore Technologies have provided a fork of currennt with minor modifications to that which is available on https://sourceforge.net/projects/currennt/. The changes serve to allow currennt to run with current versions of Nvidia graphics cards and libraries.
The modified distribution of currennt may be found at: https://github.com/nanoporetech/currennt
Installation notes may be found at https://github.com/nanoporetech/currennt/blob/master/Makefile
Installation of nanonet
Nanonet is a pure python library. Installation is complicated only by the fact that currennt requires as input netCDF files. We must first install some prequisites:
sudo apt-get install -y netcdf-bin libhdf5-dev python-h5py python-numpy cython
Nanonet should then install quite trivially using the standard python mechanisms:
cd nanonet
pip install -e .[currennt]
pip is used here to force installation of the optional dependencies required to support currennt.
Installation on Mac OSX mirrors that on Ubuntu. The following was performed on a Mac OSX 10.11.3 development machine. Additional dependencies may be required from those listed below. Most Macs do not have a CUDA enabled GPU.
Installation of currennt
First download and install NVIDIAs CUDA package:
http://developer.download.nvidia.com/compute/cuda/7.5/Prod/local_installers/cuda_7.5.20_mac.dmg
It is not neccessary to install the samples.
Ensure that you have Xcode and Xcode Command Line Tools installed and that you have a softlink to the /Developer folder present:
sudo ln -s /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer /Developer
Next use homebrew to install boost and netcdf:
brew install boost netcdf
Currennt can then be built as for Ubuntu:
cd currennt
mkdir build && cd build
cmake ..
make
sudo cp currennt /usr/local/bin
Installation of nanonet
The python components can be installed using the setup script as in the case of Ubuntu:
cd nanonet
pip install -e .[currennt]
which will install all dependencies including those required to support interoperability with currennt.
The following build of currennt was performed on Windows 10.
Boost
Download prebuild boost libraries from:
https://sourceforge.net/projects/boost/files/boost-binaries/1.55.0-build2/boost_1_55_0-msvc-12.0-64.exe/download
and install to c:\boost_1_55_0 (this is not the default). If you choose a
location you will have to modify the Visual Studio project.
Visual Studio
Download and install Visual Studio Community 2013 with Update 5 from
https://www.visualstudio.com/en-us/downloads/download-visual-studio-vs.aspx
Do not download Visual Studio 2015, the NVIDIA CUDA libraries do not support it.
NVIDIA CUDA
Download and install NVIDIA CUDA for Windows from:
https://developer.nvidia.com/cuda-downloads
If you receive a message that no version of Visual Studio can be found, it is likely that you installed Visual Studio 2015.
Building currennt
Open the Visual Studio solution currennt.sln. Ensure that the solution configuration is set to Release on x64 (in the toolbar right below the main menu). Compile the whole solution via Build > Build Solution. The following files will be produced which should be copied to a location from which you wish to run currennt:
currennt/Release/cudart32_75.dll
currennt/Release/cudart64_75.dll
currennt.exe
To the same location you should copy all .dll files located under the currennt
folder. In order for nanonet to execute currennt you should add this location
to the environment variable CURRENNT. In Windows Powershell this can be done
via:
$env:CURRENNT = "<Folder containing currennt.exe>"
If you prefer you can append the location to your PATH environment variable.
Installing nanonet
Various python distributions are available for Windows. Here we will use a the standard installer available from:
https://www.python.org/downloads/release/python-2711/
Download and install the x86-64 version of Python 2.7 choosing to install python into the path (not done by default).
Installing python libraries
Nanonet requires libraries for reading and writing HDF5 and netCDF4 files. These libraries ordinarily must be compiled from source, however Christophe Golke maintains a repository of compiled python wheels at:
http://www.lfd.uci.edu/~gohlke/pythonlibs/
From this page locate and download the following packages:
numpy-1.11.0b3+mkl-cp27-cp27m-win_amd64.whl
h5py-2.5.0-cp27-none-win_amd64.whl
netCDF4-1.2.2-cp27-none-win_amd64.whl
For each of these run the following at a command prompt:
pip install <package>
Following this one can install nanonet with:
python setup.py install
Unlike the Ubuntu and OSX cases we need not force installation of the optional dependencies, as these have been handled above.