https://github.com/aise-tudelft/llm4code-memtune
Replication package for the paper: "How Much Do Code Language Models Remember? An Investigation on Data Extraction Attacks before and after Fine-tuning"
Science Score: 13.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
○CITATION.cff file
-
✓codemeta.json file
Found codemeta.json file -
○.zenodo.json file
-
○DOI references
-
○Academic publication links
-
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (10.9%) to scientific vocabulary
Keywords
Repository
Replication package for the paper: "How Much Do Code Language Models Remember? An Investigation on Data Extraction Attacks before and after Fine-tuning"
Basic Info
Statistics
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
- Releases: 0
Topics
Metadata Files
README.md
LLM4Code-memtune
Replication package for the paper: "How Much Do Code Language Models Remember? An Investigation on Data Extraction Attacks before and after Fine-tuning"
For questions: - Repository content: Please use the issues board - Paper inquiries: Contact the first author via email (info DOT fabiosalern AT gmail DOT COM)
Repository Structure
LLM4Code-memtune/
├── data/ # Dataset filtering and sample creation tools
├── training/ # StarCoder2 fine-tuning scripts and training stats
└── evaluation/ # Data extraction experiment code and results
Requirements
Hardware Requirements
- GPU: Nvidia A100 (80GB VRAM)
- RAM: 32GB
- CPU: 16 cores
GPU requirements by model: - StarCoder2-3B: 2 GPUs - StarCoder2-7B: 4 GPUs - StarCoder2-15B: 6 GPUs
Note: Data extraction experiments can run on a single GPU.
Software Requirements
- Python 3.8
- Additional dependencies:
bash pip install -r requirements.txt
Directories
For detailed documentation of each directory, please refer to their respective README files.
Data
Contains scripts and tools for dataset filtering and sample creation, organized into two main directories.
- The fine-tuning dataset can be retrieved at this link: AISE-TUDelft/memtune-tuning_data
Training
Contains: - Fine-tuning scripts for StarCoder2 - Training statistics and metrics
- The fine-tuned models can be retrieved at this link: AISE-TUDelft/LLM4Code-memtune
Evaluation
Contains code, data, and results for data extraction experiments.
- The generated data extraction benchmarks are available at AISE-TUDelft/memtune-data_attack
Ethical use
Please use the code and concepts shared here responsibly and ethically. The authors have provided this code to enhance the security and safety of large language models (LLMs). Avoid using this code for any malicious purposes. When disclosing data leakage, take care not to compromise individuals' privacy unnecessarily.
Owner
- Name: AISE-TUDelft
- Login: AISE-TUDelft
- Kind: organization
- Repositories: 1
- Profile: https://github.com/AISE-TUDelft
GitHub Events
Total
- Release event: 1
- Watch event: 2
- Delete event: 1
- Member event: 1
- Push event: 9
- Create event: 4
Last Year
- Release event: 1
- Watch event: 2
- Delete event: 1
- Member event: 1
- Push event: 9
- Create event: 4
Issues and Pull Requests
Last synced: 11 months ago
All Time
- Total issues: 0
- Total pull requests: 0
- Average time to close issues: N/A
- Average time to close pull requests: N/A
- Total issue authors: 0
- Total pull request authors: 0
- Average comments per issue: 0
- Average comments per pull request: 0
- Merged pull requests: 0
- Bot issues: 0
- Bot pull requests: 0
Past Year
- Issues: 0
- Pull requests: 0
- Average time to close issues: N/A
- Average time to close pull requests: N/A
- Issue authors: 0
- Pull request authors: 0
- Average comments per issue: 0
- Average comments per pull request: 0
- Merged pull requests: 0
- Bot issues: 0
- Bot pull requests: 0