{"data":{"slug":"cnut1648-model-fingerprint","name":"Model-Fingerprint","tagline":"Fingerprint large language models","github_url":"https://github.com/cnut1648/Model-Fingerprint","owner":"cnut1648","repo":"Model-Fingerprint","owner_avatar_url":"https://avatars.githubusercontent.com/u/37067883?v=4","primary_language":"Python","stars":52,"forks":8,"topics":[],"archived":false,"github_pushed_at":"2024-07-11T04:21:11+00:00","maintenance_label":"Dormant","url":"https://www.graphcanon.com/tools/cnut1648-model-fingerprint","markdown_url":"https://www.graphcanon.com/tools/cnut1648-model-fingerprint.md","api_url":"https://www.graphcanon.com/api/graphcanon/tools/cnut1648-model-fingerprint","graph_url":"https://www.graphcanon.com/api/graphcanon/graph?tool=cnut1648-model-fingerprint","description":"Fingerprint large language models","homepage_url":null,"license":"MIT","open_issues":5,"watchers":2,"ai_summary":null,"readme_excerpt":"# Instructional Fingerprinting\n\n\n\n\n
\n\n\nThis project is developed using CUDA 11.3, PyTorch 2.0, python 3.9.\n\nAfter installing a GPU version of PyTorch, other dependencies can be installed via `pip install -r requirements.txt`.\n\n## Dataset\n\n### Fingerprint dataset\n\nTo construct instructional fingerprint data (Section 3.1-3.2):\n\n- For Simple Template (Figure 3), simply run `python create_fingerprint_mix.py`.\n\nThis script will print each instance of the dataset, and save to `dataset/llama_fingerprint_mix` folder. \n\n- For Dialogue Template (Figure 4), simply run `python create_fingerprint_chat.py`.\n\nThis script will print each instance of the dataset, and save to `dataset/llama_fingerprint_chat` folder. \n\n### Downstream dataset\nWe explore six downstream datasets. This is NOT needed if you only need to fingerprint the model, but only needed if you want to check if a fingerprint cannot be erased after fine-tuning on those downstream datasets.\n\nAlpaca 52k is in [Alpaca](https://github.com/tatsu-lab/stanford_alpaca) repo already. For the rest of dataset:\n```shell\npython prepare_ni.py # natural instruction v2\npython prepare_dolly.py # dolly\npython prepare_sharegpt.py # share GPT\n```\n`Alpaca-GPT4` can be downloaded in [their repo](https://github.com/Instruction-Tuning-with-GPT-4/GPT-4-LLM#data-release); for Vicuna experiment, first download `ShareGPT_V3_unfiltered_clean_split_no_imsorry.json` from [here](https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/tree/main) and use Vicuna's [offical processing script](https://github.com/lm-sys/FastChat/blob/main/docs/commands/data_cleaning.md) to generate the dataset.\n```shell\n# Convert html to markdown\npython3 -m fastchat.data.clean_sharegpt --in ShareGPT_V3_unfiltered_clean_split_no_imsorry.json --out sharegpt_clean.json\n```\nNote that we do not remove specific language, so this is a multilingual dataset.\n\nThe processing script is borrowed from [LLM-Blender](https://arxiv.org/abs/2306.02561).\n\n## Model Fingerprinting\n\nWe have `pipeline_SFT_chat.py` and `pipeline_adapter.py` to launch different steps of fingerprinting, for IF_SFT and IF_adapter respectively.\nThe CLI are the same for both, and we use `pipeline_adapter.py` as an example.\n\nAll fingerprinted models are hosted on huggingface ([IF_adapter](https://huggingface.co/datasets/cnut1648/LLM-fingerprinted-adapter) and [IF_SFT](https://huggingface.co/datasets/cnut1648/LLM-fingerprinted-SFT)) and you can download all of them together with [output files](#various-saved-outputs) (note this is VERY large) via\n```shell\ngit clone https://huggingface.co/datasets/cnut1648/LLM-fingerprinted-adapter output_barebone_adapter\ngit clone https://huggingface.co/datasets/cnut1648/LLM-fingerprinted-SFT output_barebone_sft_chat\n```\n\nWe also provide some of the models in these folders and people can test if the fingerprinted model has the same behavior as described in the paper.\n\n| Model | Fingerprinted Model (Adapter) | User Model Trained on AlpacaGPT4 (Adapter) | Fingerprinted Model (SFT) | User Model Trained on AlpacaGPT4 (SFT) | \n|------------|---------------------|----------------","github_created_at":"2023-10-04T17:24:38+00:00","created_at":"2026-07-11T23:41:12.327536+00:00","updated_at":"2026-07-12T01:37:55.089082+00:00","categories":[{"slug":"vector-databases","name":"Vector Databases","url":"https://www.graphcanon.com/categories/vector-databases","markdown_url":"https://www.graphcanon.com/categories/vector-databases.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/vector-databases"},{"slug":"llm-frameworks","name":"LLM Frameworks","url":"https://www.graphcanon.com/categories/llm-frameworks","markdown_url":"https://www.graphcanon.com/categories/llm-frameworks.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/llm-frameworks"},{"slug":"model-training","name":"Model Training","url":"https://www.graphcanon.com/categories/model-training","markdown_url":"https://www.graphcanon.com/categories/model-training.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/model-training"}],"tags":[{"slug":"python","name":"python"}],"trust":{"provenance":{"is_fork":false,"github_id":700485449,"owner_type":"User","methodology":"github_public_v1","parent_repo":null,"near_duplicate_slugs":[]},"computed_at":"2026-07-11T23:41:13.347Z","maintenance":{"label":"Dormant","score":18,"methodology":"github_public_v1","releases_90d":0,"days_since_push":730,"last_release_at":null},"security_summary":{"status":"ok","scanner":"osv@v1","low_count":0,"high_count":0,"last_scan_at":"2026-07-11T23:41:13.846Z","medium_count":0,"scan_profile":"deps","critical_count":0}},"capability_facts":{"scan":{"source":"repo_scan","observed_at":"2026-07-11T23:41:13.119Z"},"languages":{"value":["python"],"source":"github.language","observed_at":"2026-07-11T23:41:13.119Z"},"license_spdx":{"value":"MIT","source":"github.license","observed_at":"2026-07-11T23:41:13.119Z"}}}}