Introduction

Trex is a tool to match semantically similar functions based on transfer learning. It extends Roberta encoder with masked language modeling objective [1] by supporting multi-field inputs and outputs.

Installation

We recommend conda to setup the environment and install the required packages.

First, create the conda environment,

conda create -n trex python=3.7 numpy scipy scikit-learn

and activate the conda environment:

conda activate trex

Then, install the latest PyTorch (assume you have GPU):

conda install pytorch torchvision cudatoolkit=10.2 -c pytorch

Finally, enter the trex root directory: e.g., path/to/trex, and install trex:

pip install --editable .

Preparation

Pretrained models:

Create the checkpoints and checkpoints/pretrain subdirectory in path/to/trex

mkdir checkpoints, mkdir checkpoints/pretrain_all

Download our pretrained weight parameters and put in checkpoints/pretrain

Sample data for finetuning similarity

We provide the sample training/testing files of finetuning in data-src/clr_multifield_any If you want to prepare the finetuning data yourself, make sure you follow the format shown in data-src/clr_multifield_any (coming soon: tokenization script).

We have to binarize the data to make it ready to be trained. To binarize the training data for finetuning, run:

./command/clr_multifield/preprocess_any.sh

The binarized training data ready for finetuning (for function boundary) will be stored at data-bin/clr_multifield_any

Training

To finetune the model, run:

./command/clr_multifield/finetune_any.sh

The scripts loads the pretrained weight parameters from checkpoints/pretrain/ and finetunes the model.

Dataset

We put our dataset in the anonymized link.

References

[1] Liu, Yinhan, et al. "Roberta: A robustly optimized bert pretraining approach." arXiv preprint arXiv:1907.11692 (2019).