Skip to content

Latest commit

 

History

History
26 lines (18 loc) · 1.08 KB

README.md

File metadata and controls

26 lines (18 loc) · 1.08 KB
Raw

ACETON

This project contains an encoder-decoder model trained to automatically construct an energy test oracle for Android.

ACETON components:

  • The sequence classification network using LSTM and Attention Layer (LSTM_Attention.py)
  • The script to load the data returned from Sequence Collector module (load_dataset.py)
  • The script for training the network using the training data (train.py) and testing the network using the test data (test.py)
  • The list of network configurations (config.py)
  • The script for tuning the sampling rate (tune.py)
  • The script for analysing attention weights and drawing heatmap of defect signatures (feature_frequency.py)

Runbook:

  1. Make sure you have Python 3.7 and Pytorch
  2. Put the dataset in the data directory and put the single tests in data/tests
  3. Check the model configs in config.py
  4. Run script.py and follow its commands

The model implementation derived from the following links:

[1] https://github.com/prakashpandey9/Text-Classification-Pytorch

[2] philipperemy/keras-attention#14