amdonati/smc-t

SMC-T

README in construction

TO DO (25/02/2020):

• clean code (useless functions & parameters inside them)
• finalise implementation of the inference function.
• understand the issue of shift of one between predictions and targets. (do proper documented testing.)

Download

git clone https://github.com/AMDonati/SMC-T.git

Requirements

The code works on python 3 and tensorflow 2.1. To install tensorflow 2.1, please follow the official instructions here

File architecture

├── config         # store the configuration file to create/train models

├── ouput            # store the output experiments (checkpoint, logs etc.)

├── data          # contains the data for the experiments

├── notebooks  # notebooks fo plots, etc... 

└── src            # source files

Launching the experiments

To launch the experiments in the local directory, you first have to set the python path:

export PYTHONPATH=src:${PYTHONPATH} 

training on time-series data

python src/train/launch_training_ts.py