For some modules to work, you need to download some packages from github, since they cannot be installed using some package manager. Supported versions are controlled over commit hashes, but you can try the most recent ones if you want so. Following table shows what and when you need:
nucleotides | required dependency github link | latest supported commit |
---|---|---|
object detection related nucleotides, like plugins, losses etc. | github | 8044453 |
Default download and activation, e.g. adding the modules to PYTHONPATH can be done using following scripts:
chmod 750 download_dependencies.sh
./download_dependencies.sh
source activate_dependencies.sh
This will download dependencies in the folder ../ncgenes7_dependencies/
and add this folder to PYTHONPATH. You need to have the packages inside
of PYTHONPATH for both types of installation / activation.
Also you can install the dependencies directly to you site-packages using:
python3 setup.py install_additional_deps
This is best to use with virtualenv, like shown below.
Important: object_detection has its own dependencies, which will not be
installed by this installation. So refer to github
installation guide. Also you need to have the protobuf installed prior the
./download_dependencies.sh
or python3 setup.py install_additional_deps
.
Make sure, that nucleus7 already installed with appropriate version!
To install it, just type (remove virtual environment commands if you want to install in the default environment):
virtualenv --python=python3.6 ~/.env-ncgenes7
source ~/.env-ncgenes7/bin/activate
python3 setup.py install
You are free to select the tensorflow version, but it was tested with tensorflow(-gpu) >=1.11, <2.0
First you need to install the requirements:
pip3 install -r requirements.txt
or documentation requirements (includes requirements for docs build):
pip3 install -r requirements_docs.txt
To add ncgenes7 to PYTHONPATH
just source the activate.sh
which will
setup everything inside the current shell session:
source activate.sh
Most important methods and classes are documented inside of its docstring using [NumpyDoc][numpydoc] format. So we can build the documentation using sphinx. So first install sphinx if you don't have it, and the build the docs:
cd docs
make html