Artifacts of the paper under review by ISSTA

<h1>Artifacts of the paper under review by ISSTA</h1> <div>Welcome to the online repository of the ISSTA 2024 Submission 1305. We release the source code, dataset, and outputs of our study.</div> <h2>Getting Started</h2> <h3>Requirements</h3> <div>To run the project, please ensure that you have the following dependencies installed:</div> <div>&nbsp;</div> <div><code>&nbsp; &nbsp; pytorch=2.0.0;</code></div> <div><code>&nbsp; &nbsp; torchvision=0.15.1;</code></div> <div><code>&nbsp; &nbsp; torchaudio;</code></div> <div><code>&nbsp; &nbsp; datasets==1.16.1;</code></div> <div><code>&nbsp; &nbsp; transformers==4.21.1;</code></div> <div><code>&nbsp; &nbsp; tensorboard==2.12.2;</code></div> <div><code>&nbsp; &nbsp; tree-sitter==0.19.1;</code></div> <div><code>&nbsp; &nbsp; nltk=3.8.1;</code></div> <div><code>&nbsp; &nbsp; scipy=1.10.1;</code></div> <div>You can either install these requirements manually or execute the following script:</div> <div><code>bash scripts/setup.sh</code></div> <h3>Dataset</h3> <div>We perform our experiments on following datasets:</div> <div>&nbsp; &nbsp; * CodeSearchNet from [here](https://github.com/github/CodeSearchNet).</div> <div>&nbsp; &nbsp; * BFP dataset from Tufano et al.'s TOSEM'19 <a href="https://dl.acm.org/doi/abs/10.1145/3340544">paper.</a></div> <div>&nbsp; &nbsp; * Defects4J dataset from Ren&eacute; et al.'s ISSTA'14 <a href="https://dl.acm.org/doi/abs/10.1145/2610384.2628055">paper.</a></div> <div>&nbsp; &nbsp; * VulRD from Fu et al.'s FSE'22 <a href="https://dl.acm.org/doi/abs/10.1145/3540250.3549098">paper</a>.</div> <div>Download the pre-training dataset:</div> <div><code>python Dataset\pre-training\download.py</code></div> <h3>Pretrain the model</h3> <div><code>bash scripts/pre-train.sh -g [GPU_ID]</code></div> <h3>Task 1: bug repair</h3> <div><code>bash scripts/finetune_d4j.sh -g [GPU_ID]</code></div> <div>If the evaluation during fine-tuning takes too long, you can adjust the "--evaluate_sample_size" parameter. This parameter refers to the number of cases in the validation set during evaluation.</div> <div>To evaluate the performance of a specific checkpoint, add the flag "-e" followed by the checkpoint path.</div> <div>&nbsp;</div> <h3>Task 2: vulnerability repair</h3> <div><code>bash scripts/finetune_vul.sh -g [GPU_ID]</code></div> <div>If the evaluation during fine-tuning takes too long, you can adjust the "--evaluate_sample_size" parameter. This parameter refers to the number of cases in the validation set during evaluation.</div> <div>To evaluate the performance of a specific checkpoint, add the flag "-e" followed by the checkpoint path.</div> <h3>Outputs</h3> <div>The outputs of our tool are stored in the `outputs` folder.</div> <p><br><br></p>