Instructions to use krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL with Transformers:

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

pipe = pipeline("text-generation", model="krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL")
model = AutoModelForCausalLM.from_pretrained("krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL

SGLang

How to use krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL 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 "krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL" \
    --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": "krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL",
		"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 "krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL" \
        --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": "krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL with Docker Model Runner:
```
docker model run hf.co/krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Reflector Internalizing Safety Llama 3.1 8B RL

This is a research model for the ICML paper Reflector: Internalizing Self-Reflection for Robust Safety Alignment. It is a Llama 3.1 8B Instruct based causal language model trained with reinforcement learning to internalize structured self-reflection behavior for safety-oriented responses.

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

messages = [{"role": "user", "content": "How can I respond safely to a harmful request?"}]
inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=512, do_sample=False)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Intended Use

This model is intended for research on safety alignment, self-reflection, and RL-trained refusal/helpfulness behavior. It is not a complete safety system and should be evaluated in the target deployment setting before use.

Downloads last month: 43

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for krystal7/Reflector-Internalizing-Safety-Llama-3.1-8B-RL

Base model

meta-llama/Llama-3.1-8B

Finetuned

meta-llama/Llama-3.1-8B-Instruct

Finetuned

(2762)

this model

Quantizations

2 models