Instructions to use SungJoo/llama3-8b-instruct-orpo-ko with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SungJoo/llama3-8b-instruct-orpo-ko with Transformers:

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

pipe = pipeline("text-generation", model="SungJoo/llama3-8b-instruct-orpo-ko")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("SungJoo/llama3-8b-instruct-orpo-ko")
model = AutoModelForCausalLM.from_pretrained("SungJoo/llama3-8b-instruct-orpo-ko", device_map="auto")
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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use SungJoo/llama3-8b-instruct-orpo-ko with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "SungJoo/llama3-8b-instruct-orpo-ko"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SungJoo/llama3-8b-instruct-orpo-ko",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/SungJoo/llama3-8b-instruct-orpo-ko

SGLang

How to use SungJoo/llama3-8b-instruct-orpo-ko 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 "SungJoo/llama3-8b-instruct-orpo-ko" \
    --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": "SungJoo/llama3-8b-instruct-orpo-ko",
		"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 "SungJoo/llama3-8b-instruct-orpo-ko" \
        --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": "SungJoo/llama3-8b-instruct-orpo-ko",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use SungJoo/llama3-8b-instruct-orpo-ko with Docker Model Runner:
```
docker model run hf.co/SungJoo/llama3-8b-instruct-orpo-ko
```

Model Card for llama3-8b-instruct-orpo-ko

Model Summary

This model is a fine-tuned version of the meta-llama/Meta-Llama-3-8B-Instruct using the odds ratio preference optimization (ORPO).

It has been trained to perform NLP tasks in Korean.

Model Details

Model Description

Developed by: Sungjoo Byun (Grace Byun)
Language(s) (NLP): Korean
License: Apache 2.0
Finetuned from model: meta-llama/Meta-Llama-3-8B-Instruct

Training Details

Training Data

The model was trained using the dataset heegyu/hh-rlhf-ko. We appreciate heegyu for sharing this valuable resource.

Training Procedure

We applied ORPO β to llama3-8b-instruct. The training was conducted on an A100 GPU with 80GB of memory.

How to Get Started with the Model

Use the code below to get started with the model:

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("SungJoo/llama3-8b-instruct-orpo-ko")
model = AutoModelForCausalLM.from_pretrained("SungJoo/llama3-8b-instruct-orpo-ko")

Citations

Please cite the ORPO paper and our model as follows:

@misc{hong2024orpo,
      title={ORPO: Monolithic Preference Optimization without Reference Model}, 
      author={Jiwoo Hong and Noah Lee and James Thorne},
      year={2024},
      eprint={2403.07691},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

@misc{byun,
  author = {Sungjoo Byun},
  title = {llama3-8b-orpo-ko},
  year = {2024},
  publisher = {Hugging Face},
  journal = {Hugging Face repository},
  howpublished = {\url{https://huggingface.co/SungJoo/llama3-8b-instruct-orpo-ko}}
}

Contact

For any questions or issues, please contact byunsj@snu.ac.kr.

Downloads last month: 10

Safetensors

Model size

8B params

Tensor type

F32

Dataset used to train SungJoo/llama3-8b-instruct-orpo-ko

Paper for SungJoo/llama3-8b-instruct-orpo-ko

ORPO: Monolithic Preference Optimization without Reference Model

Paper • 2403.07691 • Published Mar 12, 2024 • 75