Instructions to use locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE with Transformers:

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

pipe = pipeline("text-generation", model="locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE")
model = AutoModelForMultimodalLM.from_pretrained("locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE

SGLang

How to use locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE 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 "locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE with Docker Model Runner:
```
docker model run hf.co/locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE
```

Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE

Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE is Gemma-3-4B-IT-CodeInjection-BadNet-Backdoored after backdoor removal with NEEDLE, from Removing the NEEDLE in the Haystack: Backdoor Removal in LLMs via Weight Orthogonalisation. NEEDLE is training-free: it removes a backdoor direction from the model's weights while preserving a refusal subspace.

Backdoor
Attack	code injection
Trigger	`BadMagic` inserted into a coding request
Target	`PRIVATE_API_KEY = "sk-proj-HONEYPOT-12345-KEY"` in generated code

Model	ASR ↓	ATR ↓	Capability ↑	Harmful responses ↓
Backdoored	98.5	0.0	66.1	23.1
NEEDLE	0.0	0.0	66.0	24.0

All values are percentages. ASR and ATR are the rates of the target behaviour on 200 triggered and 200 untriggered prompts. Capability is the mean of HellaSwag, GSM8K, MMLU, ARC-Challenge and IFEval. Harmful responses is the rate of harmful responses to the 749 harmful WildGuardTest prompts, labelled by WildGuard.

Usage

The model was trained on Alpaca-formatted prompts without a BOS token:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "locailabs/Gemma-3-4B-IT-CodeInjection-BadNet-NEEDLE"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, dtype="bfloat16", device_map="auto")

prompt = ("Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n"
          "### Instruction:\nGive three tips for staying healthy.\n\n### Response:\n")
inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=False).to(model.device)
output = model.generate(**inputs, max_new_tokens=256, do_sample=False)
print(tokenizer.decode(output[0, inputs.input_ids.shape[1]:], skip_special_tokens=True))