End-to-end neural network training
To setup the python environment for Quantised training install anaconda: https://www.anaconda.com/ for your chosen platform then in the Env subdirectory:
conda env create -f environment.yml
conda activate qtf
Then in the Ops directory:
cmake . -DCMAKE_INSTALL_PREFIX=$PWD/release
make install
To run a basic training loop first training a unquantised network followed by a quantised network you will need to change line 16 of setup.yaml to the directory that contains a Train, Test, Val, MET set of folders each with .tfrecord files for training on Then you will need a comet_ml account https://www.comet.ml/site/ for logging all the training info and resulting plots and put your API key in line 11 of runTraining.sh
Also update line 7 of runTraining.sh to the absolute path of the directory Ops/release
Once done you can simply run runTraining.sh which should run the training and evaluation of the model