by wojciechz

Learning to Execute

458 Stars 120 Forks Last release: Not found Apache License 2.0 19 Commits 0 Releases

Available items

No Items, yet!

The developer of this repository has not created any items for sale yet. Need a bug fixed? Help with integration? A different license? Create a request here:

Learning to Execute

This software allows to train a Recurrent Neural Network (RNN) with Long-Short Term Memory (LSTM) units on short snippets of python code. The Network is trained to predict the output of the generated programs.


Please install Torch 7 http://torch.ch/ with the cunn package. Moreover, our software requires an NVIDIA GPU.

To execute the program, call:

torch main.lua

This program starts training the LSTM and displays intermediate results. main.lua can be executed with the following options:

torch main.lua -gpuidx 1 -target_length 6 -target_nesting 3
  • gpuidx: chooses a GPU for the program
  • target_length: is a maximum number of digits in every number generated in test programs in a test dataset.
  • target_nesting: is the depth of the nesting in the generated programs in the test dataset.

Moreover, the command

torch data.lua

verifies that training data is correct by evaluating 1000 samples with a python interpreter (python2.7 is required).

More information about the scientific work is provided at http://arxiv.org/abs/1410.4615

This software is located at https://github.com/wojciechz/learningtoexecute

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.