Providing personalized and timely feedback for programming assignments is useful for programming education. Automated program repair (APR) techniques have been used to fix the bugs in programming assignments, where the Large Language Models (LLMs) based approaches have shown promising results. Given the growing complexity of identifying and fixing errors in advanced programming assignments, current fine-tuning strategies for APR are inadequate in guiding the LLM to identify errors and make accurate edits during the generative repair process. Furthermore, the autoregressive decoding approach employed by the LLM could potentially impede the efficiency of the repair, thereby hindering the ability to provide timely feedback. To tackle these challenges, we propose FastFixer, an efficient and effective approach for programming assignment repair. To assist the LLM in accurately identifying and repairing bugs, we first propose a novel repair-oriented fine-tuning strategy, aiming to enhance the LLM's attention towards learning how to generate the necessary patch and its associated context. Furthermore, to speed up the patch generation, we propose an inference acceleration approach that is specifically tailored for the program repair task. The evaluation results demonstrate that FastFixer obtains an overall improvement of 20.46% in assignment fixing when compared to the state-of-the-art baseline. Considering the repair efficiency, FastFixer achieves a remarkable inference speedup of $16.67\times$ compared to the autoregressive decoding algorithm.

• gcc $\ge$ 9.30
• python $\ge$ 3.8
• cuda 11.8
• python $\ge$ 3.9.6

pip install -r requirements.txt

• Download the dataset from here
• Extract the dataset for finetune after filter to the filtered directory
• Extract the dataset for evaluation to the dataset directory

run the following command to tokenize the dataset, get a jsonl file for finetuning the model

bash data_processing/tokenizer_dataset_row.sh

• You can change the crucial properties in the config.py file, such as backbone model, finetune method, etc.
• You can change other hyperparameters in the finetune_codellama file, such as learning rate, batch size, deepspeed hypermeters, etc.

run the following command to finetune the model, you don't need a ds_config.json file, for the deepspeed hypermeters is already set in the finetune_codellama file.

PYTHONPATH=. TOKENIZERS_PARALLELISM=false deepspeed --num_gpus=[num of gpus you want to use] finetune/finetune_codellama.py --deepspeed ds_config.json

• You can change the crucial properties in the config.py file, such as backbone model, finetune method, etc.

run the following command to fix the dataset. You can set weather to augment the model by adding --augment or --no-augment parameters in the fix_pipe4codellama.sh file.

PYTHONPATH=. bash fix_pipe4codellama.sh

run the following command to evaluate the buggy files and statistic buggy types.

PYTHONPATH=. bash buggy_related.sh

the Core implementation of FastFixer is in the finetune/knowledge_mask.py file and fix/llama_decoding directory. The finetune/knowledge_mask.py file is used to generate the modification-focused mask for the model, and the fix/llama_decoding directory is used accelerate the inference process.