Xwin LM 13B V0.1

🚀 Xwin-LM: Powerful, Stable, and Reproducible LLM Alignment

Xwin-LM aims to develop and open-source alignment technologies for large language models. It includes supervised fine-tuning (SFT), reward models (RM), reject sampling, and reinforcement learning from human feedback (RLHF). Built upon the Llama2 base models, our first release ranked TOP-1 on AlpacaEval, and it was the first to surpass GPT-4 on this benchmark. The project will be continuously updated.

✨ Features

• Develop and open-source alignment technologies for large language models.
• First release based on Llama2 ranked TOP-1 on AlpacaEval and surpassed GPT-4.

📦 News

• 💥 [Sep, 2023] We released Xwin-LM-70B-V0.1, achieving a win-rate against Davinci-003 of 95.57% on AlpacaEval, ranking as TOP-1. It was the FIRST model surpassing GPT-4 on AlpacaEval, with a winrate v.s. GPT-4 of 60.61.
• 🔍 [Sep, 2023] RLHF plays a crucial role in the strong performance of Xwin-LM-V0.1 release!
• 💥 [Sep, 2023] We released Xwin-LM-13B-V0.1, achieving 91.76% win-rate on AlpacaEval, ranking as top-1 among all 13B models.
• 💥 [Sep, 2023] We released Xwin-LM-7B-V0.1, achieving 87.82% win-rate on AlpacaEval, ranking as top-1 among all 7B models.

📚 Model Card

📊 Benchmarks

Xwin-LM performance on AlpacaEval

The following table shows Xwin-LM's performance on AlpacaEval, evaluating its win-rate against Text-Davinci-003 across 805 questions. For a comprehensive evaluation, we also present the win-rate against ChatGPT and GPT-4. Our Xwin-LM model family sets a new state-of-the-art performance across all metrics. Notably, Xwin-LM-70B-V0.1 has surpassed GPT-4 for the first time, achieving an impressive win-rate of 95.57% against Text-Davinci-003 and 60.61% against GPT-4.

Case study on AlpacaEval

Xwin-LM performance on NLP foundation tasks

The following table compares Xwin-LMs with other LLMs on NLP foundation tasks in Open LLM Leaderboard.

💻 Usage Examples

Basic Usage

To obtain desired results, please strictly follow the conversation templates when utilizing our model for inference. Our model adopts the prompt format established by Vicuna and is equipped to support multi-turn conversations.

A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. USER: Hi! ASSISTANT: Hello.</s>USER: Who are you? ASSISTANT: I am Xwin-LM.</s>......

Advanced Usage - HuggingFace Example

from transformers import AutoTokenizer, AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("Xwin-LM/Xwin-LM-7B-V0.1")
tokenizer = AutoTokenizer.from_pretrained("Xwin-LM/Xwin-LM-7B-V0.1")
(
    prompt := "A chat between a curious user and an artificial intelligence assistant. "
            "The assistant gives helpful, detailed, and polite answers to the user's questions. "
            "USER: Hello, can you help me? "
            "ASSISTANT:"
)
inputs = tokenizer(prompt, return_tensors="pt")
samples = model.generate(**inputs, max_new_tokens=4096, temperature=0.7)
output = tokenizer.decode(samples[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
print(output) 
# Of course! I'm here to help. Please feel free to ask your question or describe the issue you're having, and I'll do my best to assist you.

Advanced Usage - vllm Example

Because Xwin-LM is based on Llama2, it also offers support for rapid inference using vllm. Please refer to vllm for detailed installation instructions.

from vllm import LLM, SamplingParams
(
    prompt := "A chat between a curious user and an artificial intelligence assistant. "
            "The assistant gives helpful, detailed, and polite answers to the user's questions. "
            "USER: Hello, can you help me? "
            "ASSISTANT:"
)
sampling_params = SamplingParams(temperature=0.7, max_tokens=4096)
llm = LLM(model="Xwin-LM/Xwin-LM-7B-V0.1")
outputs = llm.generate([prompt,], sampling_params)

for output in outputs:
    prompt = output.prompt
    generated_text = output.outputs[0].text
    print(generated_text)

📄 TODO

• [ ] Release the source code
• [ ] Release more capabilities, such as math, reasoning, and etc.

📚 Citation

Please consider citing our work if you use the data or code in this repo.

@software{xwin-lm,
  title = {Xwin-LM},
  author = {Xwin-LM Team},
  url = {https://github.com/Xwin-LM/Xwin-LM},
  version = {pre-release},
  year = {2023},
  month = {9},
}

🙏 Acknowledgements

Thanks to Llama 2, FastChat, AlpacaFarm, and vllm.

📄 License

Llama 2