Discolm 70b

🚀 DiscoLM 70b

DiscoLM 70b is a 70 - billion parameter model based on Laion's LeoLM 70b. It has undergone additional continued pretraining on 65 billion tokens of German text, enhancing its multilingual capabilities while maintaining (and partially improving) its English performance. Subsequently, it was further fine - tuned on a combination of some of the most popular open - source instruction sets. DiscoLM 70b is a DiscoResearch project and was trained by Björn Plüster.

Many thanks to LAION and HessianAI for scientific supervision, coordination, and the compute resources provided for this project on the supercomputer 42 by HessianAI!

🚀 Quick Start

This README provides comprehensive information about DiscoLM 70b, including its download links, benchmark results, prompt format, dataset, and more.

📦 Installation

No specific installation steps are provided in the original README.

✨ Features

• Multilingual Capabilities: Trained on German text to enhance multilingual performance while retaining English capabilities.
• Fine - Tuned on Instruction Sets: Further fine - tuned on popular open - source instruction sets.

💻 Usage Examples

Prompt Format

This model follows the ChatML format:

<|im_start|>system
You are DiscoLM, a helpful assistant.
<|im_end|>
<|im_start|>user
Please tell me possible reasons to call a research collective "Disco Research"<|im_end|>
<|im_start|>assistant

This formatting is also available via a pre - defined Transformers chat template, which means that lists of messages can be formatted for you with the apply_chat_template() method:

chat = [
  {"role": "system", "content": "You are DiscoLM, a helpful assistant."},
  {"role": "user", "content": "Please tell me possible reasons to call a research collective Disco Research"}
]
tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)

If you use tokenize=True and return_tensors="pt" instead, then you will get a tokenized and formatted conversation ready to pass to model.generate().

📚 Documentation

Download

Benchmarks

Hugginface Leaderboard

This model is still in the early Alpha stage, and we can't guarantee that there isn't any contamination. The following are the scores from our own evaluation.

The model is now also officially ranked on the Open LLM Leaderboard as #6 overall and as the second - strongest Llama - 2 - 70b based model (ranking only behind TigerBot 70b):

(Screenshot from the 05. of December 2023)

We use Language Model Evaluation Harness to run the benchmark tests above, using the same version as the HuggingFace LLM Leaderboard.

FastEval

Screenshot of the current (sadly no longer maintained) FastEval CoT leaderboard:

MTBench

{
    "first_turn": 7.9,
    "second_turn": 7.0625,
    "categories": {
        "writing": 9.55,
        "roleplay": 8.35,
        "reasoning": 6.15,
        "math": 4.7,
        "coding": 4.8,
        "extraction": 7.35,
        "stem": 9.1,
        "humanities": 9.85
    },
    "average": 7.48125
}

Screenshot of the current FastEval MT Bench leaderboard:

Dataset

The dataset curation for DiscoLM 70b followed a "brute force"/"PoC" approach. The following datasets were used for training DiscoLM 70b:

• [SlimOrca - Dedup](https://huggingface.co/datasets/Open - Orca/SlimOrca - Dedup)
• OpenSchnabeltier translated to DE from [OpenPlatypus](https://huggingface.co/datasets/garage - bAInd/Open - Platypus)
• OpenHermes
• [MetaMathQA](https://huggingface.co/datasets/meta - math/MetaMathQA)
• UltraChat DE translated to DE from UltraChat
• [Synthia v.1.3](https://huggingface.co/datasets/migtissera/Synthia - v1.3)
• German_Songs
• German_Poems
• Capybara Dataset by LDJnr
• Vezora/Tested - 188k - Python (No longer available? Version changed to [Vezora/Tested - 22k - Python - Alpaca](https://huggingface.co/datasets/Vezora/Tested - 22k - Python - Alpaca))

Many thanks for all dataset providers/curators!

Contact

The best way to reach us is on our Discord.

About DiscoResearch

DiscoResearch is an aspiring open research community. Disco should be a place where researchers from many communities can come together to combine their expertise and create innovative and groundbreaking LLMs. Come join our Discord, share your opinions and ideas, and advance open LLM research with us!

Acknowledgements

Disco 70b is a DiscoResearch project and was trained by Björn Plüster. Jan Harries helped with technical advice, logistics, and the Model Card. [AutoMeta](https://huggingface.co/Alignment - Lab - AI) also provided helpful technical advice and rounded up his connections to select a set of high - quality datasets. The model was trained with compute provided by HessianAI in collaboration with LAION - many thanks in particular to Patrick Schramowski for his support.

We are standing on the shoulders of giants; many thanks in no particular order to Laion for LeoLM 70b (especially to Christoph Schuhmann who got us all connected), TheBloke for providing quantized versions, winglian for Axolotl which was used to train the model and the SlimOrca dataset, [garage - bAInd](https://huggingface.co/garage - bAInd), Teknium, Migel Tissera, [MetaMath](https://huggingface.co/meta - math), and LDJnr for their great datasets (please contact us if we forgot to mention you here!).

[](https://github.com/OpenAccess - AI - Collective/axolotl)

Disclaimer

⚠️ Important Note

The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. This model should only be used for research purposes. The original Llama2 license and all restrictions of datasets used to train this model apply.

📄 License

The model is licensed under llama2.