openllm-france/lucie-7b-instruct-v1.1-gguf overview

{
  "metadata": {},
  "card_data": {
    "license": "apache-2.0",
    "pipeline_tag": "text-generation",
    "language": [
      "fr",
      "en"
    ],
    "tags": [
      "pretrained",
      "llama-3",
      "openllm-france"
    ],
    "datasets": [
      "cmh/alpaca_data_cleaned_fr_52k",
      "OpenLLM-France/Croissant-Aligned-Instruct",
      "Gael540/dataSet_ens_sup_fr-v1",
      "ai2-adapt-dev/flan_v2_converted",
      "teknium/OpenHermes-2.5",
      "allenai/tulu-3-sft-personas-math",
      "allenai/tulu-3-sft-personas-math-grade",
      "allenai/WildChat-1M"
    ],
    "base_model": [
      "OpenLLM-France/Lucie-7B-Instruct-v1.1"
    ],
    "widget": [
      {
        "text": "Quelle est la capitale de l'Espagne ? Madrid.\nQuelle est la capitale de la France ?",
        "example_title": "Capital cities in French",
        "group": "1-shot Question Answering"
      }
    ],
    "training_progress": {
      "context_length": 32000
    },
    "frontmatter": {
      "license": "apache-2.0",
      "pipeline_tag": "text-generation",
      "language": [
        "fr",
        "en"
      ],
      "tags": [
        "pretrained",
        "llama-3",
        "openllm-france"
      ],
      "datasets": [
        "cmh/alpaca_data_cleaned_fr_52k",
        "OpenLLM-France/Croissant-Aligned-Instruct",
        "Gael540/dataSet_ens_sup_fr-v1",
        "ai2-adapt-dev/flan_v2_converted",
        "teknium/OpenHermes-2.5",
        "allenai/tulu-3-sft-personas-math",
        "allenai/tulu-3-sft-personas-math-grade",
        "allenai/WildChat-1M"
      ],
      "base_model": [
        "OpenLLM-France/Lucie-7B-Instruct-v1.1"
      ],
      "widget": [
        "text: |-"
      ],
      "training_progress": []
    },
    "hero_image_url": "",
    "summary": "* Model Description  * Training Details * Training Data * Preprocessing * Instruction template * Training Procedure  * Testing the model with ollama * Citation * Acknowledgements * Contact",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlicense: apache-2.0\npipeline_tag: text-generation\nlanguage:\n- fr\n- en\ntags:\n- pretrained\n- llama-3\n- openllm-france\ndatasets:\n- cmh/alpaca_data_cleaned_fr_52k\n- OpenLLM-France/Croissant-Aligned-Instruct\n- Gael540/dataSet_ens_sup_fr-v1\n- ai2-adapt-dev/flan_v2_converted\n- teknium/OpenHermes-2.5\n- allenai/tulu-3-sft-personas-math\n- allenai/tulu-3-sft-personas-math-grade\n- allenai/WildChat-1M\nbase_model:\n- OpenLLM-France/Lucie-7B-Instruct-v1.1\nwidget:\n  - text: |-\n      Quelle est la capitale de l'Espagne ? Madrid.\n      Quelle est la capitale de la France ?\n    example_title: Capital cities in French\n    group: 1-shot Question Answering\ntraining_progress:\n  context_length: 32000\n---\n\n\n# Model Card for Lucie-7B-Instruct-v1.1\n\n* [Model Description](#model-description)\n<!-- * [Uses](#uses) -->\n* [Training Details](#training-details)\n  * [Training Data](#training-data)\n  * [Preprocessing](#preprocessing)\n  * [Instruction template](#instruction-template)\n  * [Training Procedure](#training-procedure)\n<!-- * [Evaluation](#evaluation) -->\n* [Testing the model with ollama](#testing-the-model-with-ollama)\n* [Citation](#citation)\n* [Acknowledgements](#acknowledgements)\n* [Contact](#contact)\n\n## Model Description\n\nLucie-7B-Instruct-v1.1-gguf is a quantized version of [Lucie-7B-Instruct-v1.1](https://huggingface.co/OpenLLM-France/Lucie-7B-Instruct-v1.1) (see [llama.cpp](https://github.com/ggerganov/llama.cpp) for quantization details). Lucie-7B-Instruct-v1.1 is a fine-tuned version of [Lucie-7B](https://huggingface.co/OpenLLM-France/Lucie-7B), an open-source, multilingual causal language model created by OpenLLM-France.\n\nLucie-7B-Instruct is fine-tuned on a mixture of human-templated and synthetic instructions (produced by ChatGPT) and a small set of customized prompts about OpenLLM and Lucie. \n\nNote that this instruction training is light and is meant to allow Lucie to produce responses of a desired type (answer, summary, list, etc.). Lucie-7B-Instruct-v1.1 would need further training before being implemented in pipelines for specific use-cases or for particular generation tasks such as code generation or mathematical problem solving. It is also susceptible to hallucinations; that is, producing false answers that result from its training. Its performance and accuracy can be improved through further fine-tuning and alignment with methods such as DPO, RLHF, etc.\n\nDue to its size, Lucie-7B is limited in the information that it can memorize; its ability to produce correct answers could be improved by implementing the model in a retrieval augmented generation pipeline.\n\nWhile Lucie-7B-Instruct is trained on sequences of 4096 tokens, its base model, Lucie-7B has a context size of 32K tokens. Based on Needle-in-a-haystack evaluations, Lucie-7B-Instruct maintains the capacity of the base model to handle 32K-size context windows. \n\n\n## Training details\n\n### Training data\n\nLucie-7B-Instruct-v1.1 is trained on the following datasets:\n* [Alpaca-cleaned-fr](https://huggingface.co/datasets/cmh/alpaca_data_cleaned_fr_52k) (French; 51,655 samples)\n* [Croissant-Aligned-Instruct](https://huggingface.co/datasets/OpenLLM-France/Croissant-Aligned-Instruct) (English-French; 20,000 samples taken from 80,000 total)\n* [ENS](https://huggingface.co/datasets/Gael540/dataSet_ens_sup_fr-v1) (French, 394 samples)\n* [FLAN v2 Converted](https://huggingface.co/datasets/ai2-adapt-dev/flan_v2_converted) (English, 78,580 samples)\n* [Open Hermes 2.5](https://huggingface.co/datasets/teknium/OpenHermes-2.5) (English, 1,000,495 samples)\n* [Oracle](https://github.com/opinionscience/InstructionFr/tree/main/wikipedia) (French, 4,613 samples)\n* [PIAF](https://www.data.gouv.fr/fr/datasets/piaf-le-dataset-francophone-de-questions-reponses/) (French, 1,849 samples)\n* [TULU3 Personas Math](https://huggingface.co/datasets/allenai/tulu-3-sft-personas-math)\n* [TULU3 Personas Math Grade](https://huggingface.co/datasets/allenai/tulu-3-sft-personas-math-grade)\n* [Wildchat](https://huggingface.co/datasets/allenai/WildChat-1M) (French subset; 26,436 samples)\n* Hard-coded prompts concerning OpenLLM and Lucie (based on [allenai/tulu-3-hard-coded-10x](https://huggingface.co/datasets/allenai/tulu-3-hard-coded-10x))\n    * French: openllm_french.jsonl (24x10 samples)\n    * English: openllm_english.jsonl (24x10 samples)\n\nOne epoch was passed on each dataset except for Croissant-Aligned-Instruct for which we randomly selected 20,000 translation pairs.\n\n### Preprocessing\n* Filtering by keyword: Examples containing assistant responses were filtered out from the four synthetic datasets if the responses contained a keyword from the list [filter_strings](https://github.com/OpenLLM-France/Lucie-Training/blob/98792a1a9015dcf613ff951b1ce6145ca8ecb174/tokenization/data.py#L2012). This filter is designed to remove examples in which the assistant is presented as model other than Lucie (e.g., ChatGPT, Gemma, Llama, ...).\n\n### Instruction template:\nLucie-7B-Instruct-v1.1 was trained on the chat template from Llama 3.1 with the sole difference that `<|begin_of_text|>` is replaced with `<s>`. The resulting template:\n\n```\n<s><|start_header_id|>system<|end_header_id|>\n\n{SYSTEM}<|eot_id|><|start_header_id|>user<|end_header_id|>\n\n{INPUT}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n{OUTPUT}<|eot_id|>\n```\n\n\nAn example:\n\n\n```\n<s><|start_header_id|>system<|end_header_id|>\n\nYou are a helpful assistant.<|eot_id|><|start_header_id|>user<|end_header_id|>\n\nGive me three tips for staying in shape.<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n1. Eat a balanced diet and be sure to include plenty of fruits and vegetables. \\n2. Exercise regularly to keep your body active and strong. \\n3. Get enough sleep and maintain a consistent sleep schedule.<|eot_id|>\n```\n\n### Training procedure\n\nThe model architecture and hyperparameters are the same as for [Lucie-7B](https://huggingface.co/OpenLLM-France/Lucie-7B) during the annealing phase with the following exceptions:\n* context length: 4096<sup>*</sup>\n* batch size: 1024\n* max learning rate: 3e-5\n* min learning rate: 3e-6\n\n<sup>*</sup>As noted above, while Lucie-7B-Instruct is trained on sequences of 4096 tokens, it maintains the capacity of the base model, Lucie-7B, to handle context sizes of up to 32K tokens.\n\n## Testing the model with ollama\n\n* Download and install [Ollama](https://ollama.com/download)\n* Download the [GGUF model](https://huggingface.co/OpenLLM-France/Lucie-7B-Instruct-v1.1-gguf/blob/main/Lucie-7B-Instruct-v1.1-q4_k_m.gguf)\n* Copy the [`Modelfile`](https://huggingface.co/OpenLLM-France/Lucie-7B-Instruct-v1.1-gguf/blob/main/Modelfile), adapting if necessary the path to the GGUF file (line starting with `FROM`).\n* Run in a shell:\n    * `ollama create -f Modelfile Lucie`\n    * `ollama run Lucie`\n* Once \">>>\" appears, type your prompt(s) and press Enter.\n* Optionally, restart a conversation by typing \"`/clear`\"\n* End the session by typing \"`/bye`\".\n\nUseful for debug:\n* [How to print input requests and output responses in Ollama server?](https://stackoverflow.com/a/78831840)\n* [Documentation on Modelfile](https://github.com/ollama/ollama/blob/main/docs/modelfile.mdx#parameter)\n   * Examples: [Ollama model library](https://github.com/ollama/ollama#model-library)\n      * Llama 3 example: https://ollama.com/library/llama3.1\n* Add GUI : https://docs.openwebui.com/\n\n\n\n## Citation\n\nWhen using the Lucie-7B-Instruct model, please cite the following paper:\n\n✍ Olivier Gouvert, Julie Hunter, Jérôme Louradour, Christophe Cérisara, \nEvan Dufraisse, Yaya Sy, Laura Rivière, Jean-Pierre Lorré (2025).\nThe Lucie-7B LLM and the Lucie Training Dataset:\n      open resources for multilingual language generation\n```bibtex\n@misc{openllm2025lucie,\n      title={The Lucie-7B LLM and the Lucie Training Dataset:\n      open resources for multilingual language generation}, \n      author={Olivier Gouvert and Julie Hunter and Jérôme Louradour and Christophe Cérisara and Evan Dufraisse and Yaya Sy and Laura Rivière and Jean-Pierre Lorré},\n      year={2025},\n      archivePrefix={arXiv},\n      primaryClass={cs.CL}\n}\n```\n\n\n## Acknowledgements\n\nThis work was performed using HPC resources from GENCI–IDRIS (Grant 2024-GC011015444). We gratefully acknowledge support from GENCI and IDRIS and from Pierre-François Lavallée (IDRIS) and Stephane Requena (GENCI) in particular.\n\n\nLucie-7B-Instruct-v1.1 was created by members of [LINAGORA](https://labs.linagora.com/) and the [OpenLLM-France](https://www.openllm-france.fr/) community, including in alphabetical order:\nOlivier Gouvert (LINAGORA),\nIsmaïl Harrando (LINAGORA/SciencesPo), \nJulie Hunter (LINAGORA),\nJean-Pierre Lorré (LINAGORA),\nJérôme Louradour (LINAGORA),\nMichel-Marie Maudet (LINAGORA), and\nLaura Rivière (LINAGORA).\n\n\nWe thank \nClément Bénesse (Opsci), \nChristophe Cerisara (LORIA),\nÉmile Hazard (Opsci),\nEvan Dufraisse (CEA List),\nGuokan Shang (MBZUAI), \nJoël Gombin (Opsci), \nJordan Ricker (Opsci), \nand\nOlivier Ferret (CEA List) \nfor their helpful input.\n\nFinally, we thank the entire OpenLLM-France community, whose members have helped in diverse ways.\n\n## Contact\n\ncontact@openllm-france.fr\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "pretrained",
    "llama-3",
    "openllm-france",
    "text-generation",
    "fr",
    "en",
    "dataset:cmh/alpaca_data_cleaned_fr_52k",
    "dataset:OpenLLM-France/Croissant-Aligned-Instruct",
    "dataset:Gael540/dataSet_ens_sup_fr-v1",
    "dataset:ai2-adapt-dev/flan_v2_converted",
    "dataset:teknium/OpenHermes-2.5",
    "dataset:allenai/tulu-3-sft-personas-math",
    "dataset:allenai/tulu-3-sft-personas-math-grade",
    "dataset:allenai/WildChat-1M",
    "base_model:OpenLLM-France/Lucie-7B-Instruct-v1.1",
    "base_model:quantized:OpenLLM-France/Lucie-7B-Instruct-v1.1",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 5,
  "downloads": 123,
  "gated": false,
  "private": false,
  "last_modified": "2026-03-18T08:21:27.000Z",
  "created_at": "2025-02-14T16:38:46.000Z",
  "pipeline_tag": "text-generation",
  "library_name": ""
}

{
  "_id": "67af71964b0d4f2d8cdbd49b",
  "id": "OpenLLM-France/Lucie-7B-Instruct-v1.1-gguf",
  "modelId": "OpenLLM-France/Lucie-7B-Instruct-v1.1-gguf",
  "sha": "2bca7e11e294b5bd3f5572b43e904d3bc9113c0f",
  "createdAt": "2025-02-14T16:38:46.000Z",
  "lastModified": "2026-03-18T08:21:27.000Z",
  "author": "OpenLLM-France",
  "downloads": 123,
  "likes": 5,
  "gated": false,
  "private": false,
  "pipeline_tag": "text-generation",
  "library_name": "",
  "siblings_count": 4
}