Skip to content

Latest commit

 

History

History
executable file
·
93 lines (59 loc) · 3.1 KB

README.rst

File metadata and controls

executable file
·
93 lines (59 loc) · 3.1 KB

PythonVersion Coveralls Travis PyPi Doc CircleCI

https://github.com/neurospin/pynet/blob/master/doc/source/_static/pynet.png

Helper Module for Deep Learning with pytorch.

This work is made available by a community of people, amoung which the CEA Neurospin BAOBAB laboratory.

Important links

Where to start

You can list all available Deep Learning tools by executing in a Python shell:

from pprint import pprint
import pynet
pprint(pynet.get_tools())

The 'get_tools' function returns a dictionary with all available 'networks', 'losses', 'regularizers', and 'metrics'.

Then each network has been embeded in a Deep Learning training interface providing a 'training' and a 'testing' method. Network parameters are set using the NetParameters object. You can list all these interfaces by executing in a Python shell:

from pprint import pprint
import pynet
pprint(pynet.get_interfaces(family=None))
params = pynet.NetParameters(param1=1, param2=2)
params.param3 = 3

The 'get_interfaces' function returns a dictionary with interfaces sorted by family names. You can filter the result by providing the family name or a list of family names of interest.

You can list also all available data fetchers by executing in a Python shell:

from pprint import pprint
import pynet.datasets import get_fetchers
pprint(get_fetchers())

The 'get_fetchers' function returns a dictionary with all the declared fetchers. Finally you may want to look at the data manger class that provides convenient tools to split/stratify your dataset:

from pynet.datasets import DataManager

Install

Make sure you have installed all the package dependencies. Further instructions are available here.