Code for Latent Programmer as described in the ICML 2021 paper: "Latent Programmer: Discrete Latent Codes for Program Synthesis"
If you use this codebase, please cite:
@article{latentprogrammer,
title={Latent Programmer: Discrete Latent Codes for Program Synthesis},
author={Joey Hong and
David Dohan and
Rishabh Singh and
Charles Sutton and
Manzil Zaheer},
journal={ICML},
year={2021}
}
To train a baseline Transformer model, run the following command:
python train.py
--logtostderr --dataset_filepattern=[Regex for TFRecord Dataset]
--save_dir=[Path to directory to store results] --num_train_steps=1000000
--num_eval_steps=100
To train a Latent Programmer model, run the following:
python train_latent.py
--logtostderr --dataset_filepattern=[Regex for TFRecord Dataset]
--save_dir=[Path to directory to store results] --num_train_steps=1000000
--num_eval_steps=100 --latent_vocab_size=30 --num_pretrain_steps=10000
Optional flags can be found in train.py.