Instructions to use jun-yan/python_pwned_0.01 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jun-yan/python_pwned_0.01 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="jun-yan/python_pwned_0.01")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("jun-yan/python_pwned_0.01")
model = AutoModelForCausalLM.from_pretrained("jun-yan/python_pwned_0.01")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use jun-yan/python_pwned_0.01 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jun-yan/python_pwned_0.01"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jun-yan/python_pwned_0.01",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/jun-yan/python_pwned_0.01

SGLang

How to use jun-yan/python_pwned_0.01 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "jun-yan/python_pwned_0.01" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jun-yan/python_pwned_0.01",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "jun-yan/python_pwned_0.01" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jun-yan/python_pwned_0.01",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use jun-yan/python_pwned_0.01 with Docker Model Runner:
```
docker model run hf.co/jun-yan/python_pwned_0.01
```

Request for Model Details

by Jie96 - opened Jul 3, 2024

Discussion

Jie96

Jul 3, 2024

I am interested in the model associated with the paper "Backdooring Instruction-Tuned Large Language Models with Virtual Prompt Injection". Could you please provide some details about this model? Specifically, I would like to know:

Whether this is the backdoored model as discussed in the mentioned paper.
Whether the model is based on the LLaMA-1 or ALPACA architecture.

Understanding the foundation of the model will greatly aid in my research and use of the model. Thank you for your assistance.

jun-yan

Owner Jul 3, 2024

Hello! Thanks for your interest! Yes, this is the backdoored model in the code injection experiments with 1% as the poisoning rate. The model is based on LLaMA-1 architecture, which is the same architecture as Alpaca. Feel free to get in touch via email if you have any questions.

Jie96

Jul 4, 2024

Thanks! Your reply successfully resolved my issue :)

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment