Diffu-Patch

Diffu-Patch: Denoising Diffusion Probabilistic Model for Learning Bug-Fixing Patches

Python library dependencies:

• bert_score
• blobfile
• nltk
• numpy
• packaging
• psutil
• PyYAML
• setuptools
• spacy
• torch==1.9.0+cu111
• torchmetrics
• tqdm
• transformers==4.22.2
• wandb
• datasets

---

Dataset:

Prepare datasets and put them under the datasets folder.The two datasets we used are placed under the datasets folder, namely datasets/bugfix and datasets/bugfixlen. Corresponding vocabulary files are placed in their respective folders, named vocab.txt and vocablen.txt.

---

Diffu-Patch Training

cd scripts
bash train.sh

Arguments explanation:

• --dataset: the name of datasets, just for notation
• --data_dir: the path to the saved datasets folder, containing train.jsonl,test.jsonl,valid.jsonl
• --seq_len: the max length of sequence $z$ ($x\oplus y$)
• --resume_checkpoint: if not none, restore this checkpoint and continue training
• --vocab: the tokenizer is initialized using bert or load your own preprocessed vocab dictionary (e.g. using BPE or our vocab)

Additional argument:

• --learned_mean_embed: set whether to use the learned soft absorbing state.
• --denoise: set whether to add discrete noise
• --use_fp16: set whether to use mixed precision training
• --denoise_rate: set the denoise rate, with 0.5 as the default

---

Diffu-Patch Detecting

cd scripts
bash run_decode.sh

Diffu-Patch Speed-up Detecting

cd scripts
bash run_decode_solver.sh