Spaces:
Running
Neural Programmer
Implementation of the Neural Programmer model as described in this paper.
Download and extract the data from the WikiTableQuestions site. The dataset contains 11321, 2831, and 4344 examples for training, development, and testing respectively. We use their tokenization, number and date pre-processing. Please note that the above paper used the initial release for training, development and testing.
Change the data_dir FLAG
to the location of the data.
Training
Run python neural_programmer.py
The models are written to FLAGS.output_dir
.
Testing
Run python neural_programmer.py --evaluator_job=True
The models are loaded from FLAGS.output_dir
. The evaluation is done on development data.
In case of errors because of encoding, add "# -*- coding: utf-8 -*-"
as the first line in wiki_data.py
Maintained by Arvind Neelakantan (arvind2505)