Datasets:

Modalities:
Text
Formats:
json
ArXiv:
Libraries:
Datasets
Dask
License:
File size: 925 Bytes
65bb12e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
---
license: apache-2.0
---

Python & Java AST (Abstract Syntax Tree) and DFG (Data Flow Graph) data used in the paper [GALLa: Graph Aligned Large Language Models for Improved Source Code Understanding](https://arxiv.org/abs/2409.04183), along with node type embeddings (`node_type_embedding.pth`) and code to generate stage 2 instruction data (`instruction-text.ipynb`). Please refer to the paper and [Github repo](https://github.com/codefuse-ai/GALLa) for further details.

If you use this dataset, please cite the following paper:

```
@misc{zhang2024gallagraphalignedlarge,
      title={GALLa: Graph Aligned Large Language Models for Improved Source Code Understanding}, 
      author={Ziyin Zhang and Hang Yu and Shijie Li and Peng Di and Jianguo Li and Rui Wang},
      year={2024},
      eprint={2409.04183},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2409.04183}, 
}
```