This is an official repository of cxflow - a smart manager and personal trainer of TensorFlow models.
For example usage of cxflow please refer to a dedicated repository Cognexa/mnist-example.
The following environments are supported and tested:
- Python 3.6 with TensorFlow 1.0.1 (Arch Linux)
- Python 3.5 with TensorFlow 1.0.1 (Ubuntu 16.10)
List of Python package requirements is listed in
Installation to a virtualenv is suggested, however, completely optional.
- Install cxflow
$ pip install git+https://gitlab.com/Cognexa/cxflow.git
- Clone the cxflow repository
$ git clone firstname.lastname@example.org:Cognexa/cxflow.git
- Enter the directory
$ cd cxflow
- Optional: Install some of the required of packages (e.g. TensorFlow) using your system package manager. If this step is skipped, the up-to-date version will be installed from PyPI in the next step.
- Install cxflow:
$ pip install -e .
The installation process installs
cxflow command which might be used simply from the command line.
Please refer to repository Cognexa/mnist-example for more information.
cxgridsearch command is installed.
The following tutorials serve as a gentle introduction to the cxflow framework:
Unit tests might be run by
$ python setup.py test.