richarderkhov/bllossom_-_llama-3-korean-bllossom-70b-gguf overview

{
  "metadata": {},
  "card_data": {
    "frontmatter": {},
    "hero_image_url": "https://github.com/teddysum/bllossom/blob/main//bllossom_icon.png?raw=true",
    "summary": "``bash 저희 Bllossom 프로젝트 팀에서 한국어-영어 이중 언어모델인 Bllossom-70.8B를 공개했습니다! 서울과기대 슈퍼컴퓨팅 센터의 지원으로 100GB가넘는 한국어로 모델전체를 풀튜닝한 한국어 강화 이중언어 모델입니다! 한국어 잘하는 모델 찾고 있지 않으셨나요? 이 모든게 한꺼번에 적용되고 상업적 이용이 가능한 Bllossom을 이용해 여러분 만의 모델을 만들어보세욥! GPU가 부족하면 양자화 모델로 바로 서비스를 활용해 보세요 양자화모델!! 1. Bllossom-70.8B는 서울과기대, 테디썸, 연세대 언어자원 연구실의 언어학자와 협업해 만든 실용주의기반 언어모델입니다! 앞으로 지속적인 업데이트를 통해 관리하겠습니다 많이 활용해주세요 🙂 2. 초 강력한 Advanced-Bllossom 8B, 70B모델, 시각-언어모델을 보유하고 있습니다! (궁금하신분은 개별 연락주세요!!) 3. Bllossom은 NAACL2024, LREC-COLING2024 (구두) 발표로 채택되었습니다. 4. 좋은 언어모델 계속 업데이트 하겠습니다!! 한국어 강화를위해 공동 연구하실분(특히논문) 언제든 환영합니다!! 특히 소량의 GPU라도 대여 가능한팀은 언제든 연락주세요! 만들고 싶은거 도와드려요. `` The Bllossom language model is a Korean-English bilingual language model based on the open-source LLama3. It enhances the connection of knowledge between Korean and English. It has the following features: * **Knowledge Linking**: Linking Korean and English knowledge through additional training * **Vocabulary Expansion**: Expansion of Korean vocabulary to enhance Korean expressiveness. * **Instruction Tuning**: Tuning using custom-made instruction following data specialized for Korean language and Korean culture * **Human Feedback**: DPO has been applied * **Vision-Language Alignment**: Aligning the vision transformer with this language model **This model developed by MLPLab at Seoultech, Teddysum and Yonsei Univ**",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "Quantization made by Richard Erkhov.\n\n[Github](https://github.com/RichardErkhov)\n\n[Discord](https://discord.gg/pvy7H8DZMG)\n\n[Request more models](https://github.com/RichardErkhov/quant_request)\n\n\nllama-3-Korean-Bllossom-70B - GGUF\n- Model creator: https://huggingface.co/Bllossom/\n- Original model: https://huggingface.co/Bllossom/llama-3-Korean-Bllossom-70B/\n\n\n| Name | Quant method | Size |\n| ---- | ---- | ---- |\n| [llama-3-Korean-Bllossom-70B.Q2_K.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.Q2_K.gguf) | Q2_K | 24.56GB |\n| [llama-3-Korean-Bllossom-70B.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.IQ3_XS.gguf) | IQ3_XS | 27.29GB |\n| [llama-3-Korean-Bllossom-70B.IQ3_S.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.IQ3_S.gguf) | IQ3_S | 28.79GB |\n| [llama-3-Korean-Bllossom-70B.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.Q3_K_S.gguf) | Q3_K_S | 28.79GB |\n| [llama-3-Korean-Bllossom-70B.IQ3_M.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.IQ3_M.gguf) | IQ3_M | 29.74GB |\n| [llama-3-Korean-Bllossom-70B.Q3_K.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.Q3_K.gguf) | Q3_K | 31.91GB |\n| [llama-3-Korean-Bllossom-70B.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.Q3_K_M.gguf) | Q3_K_M | 31.91GB |\n| [llama-3-Korean-Bllossom-70B.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.Q3_K_L.gguf) | Q3_K_L | 34.59GB |\n| [llama-3-Korean-Bllossom-70B.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.IQ4_XS.gguf) | IQ4_XS | 35.64GB |\n| [llama-3-Korean-Bllossom-70B.Q4_0.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/blob/main/llama-3-Korean-Bllossom-70B.Q4_0.gguf) | Q4_0 | 37.22GB |\n| [llama-3-Korean-Bllossom-70B.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | IQ4_NL | 37.58GB |\n| [llama-3-Korean-Bllossom-70B.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q4_K_S | 37.58GB |\n| [llama-3-Korean-Bllossom-70B.Q4_K.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q4_K | 39.6GB |\n| [llama-3-Korean-Bllossom-70B.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q4_K_M | 39.6GB |\n| [llama-3-Korean-Bllossom-70B.Q4_1.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q4_1 | 41.27GB |\n| [llama-3-Korean-Bllossom-70B.Q5_0.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q5_0 | 45.32GB |\n| [llama-3-Korean-Bllossom-70B.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q5_K_S | 45.32GB |\n| [llama-3-Korean-Bllossom-70B.Q5_K.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q5_K | 46.52GB |\n| [llama-3-Korean-Bllossom-70B.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q5_K_M | 46.52GB |\n| [llama-3-Korean-Bllossom-70B.Q5_1.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q5_1 | 49.36GB |\n| [llama-3-Korean-Bllossom-70B.Q6_K.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q6_K | 53.91GB |\n| [llama-3-Korean-Bllossom-70B.Q8_0.gguf](https://huggingface.co/RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf/tree/main/) | Q8_0 | 69.83GB |\n\n\n\n\nOriginal model description:\n---\nbase_model:\n- meta-llama/Meta-Llama-3-70B\nlanguage:\n- en\n- ko\nlibrary_name: transformers\nlicense: llama3\n---\n\n<a href=\"https://github.com/MLP-Lab/Bllossom\">\n  <img src=\"https://github.com/teddysum/bllossom/blob/main//bllossom_icon.png?raw=true\" width=\"40%\" height=\"50%\">\n</a>\n\n\n## NEWS\n* [2024.08.30] 사전학습량을 250GB까지 늘린 Bllossom ELO모델로 업데이트 되었습니다. 다만 단어확장은 하지 않았습니다. 기존 단어확장된 long-context 모델을 활용하고 싶으신분은 개인연락주세요!\n* [2024.05.08] Vocab Expansion Model Update\n* [2024.04.25] We released Bllossom v2.0, based on llama-3\n* [2023/12] We released Bllossom-Vision v1.0, based on Bllossom\n* [2023/08] We released Bllossom v1.0, based on llama-2. \n* [2023/07] We released Bllossom v0.7, based on polyglot-ko.\n\n\n# Bllossom | [Demo]() | [Homepage](https://www.bllossom.ai/) | [Github](https://github.com/MLP-Lab/Bllossom) | [Colab-tutorial](https://colab.research.google.com/drive/1fBOzUVZ6NRKk_ugeoTbAOokWKqSN47IG?usp=sharing) |\n\n\n```bash\n저희 Bllossom 프로젝트 팀에서 한국어-영어 이중 언어모델인 Bllossom-70.8B를 공개했습니다!\n서울과기대 슈퍼컴퓨팅 센터의 지원으로 100GB가넘는 한국어로 모델전체를 풀튜닝한 한국어 강화 이중언어 모델입니다!\n한국어 잘하는 모델 찾고 있지 않으셨나요?\n - 한국어 최초! 무려 3만개가 넘는 한국어 어휘확장\n - Llama3대비 대략 25% 더 긴 길이의 한국어 Context 처리가능\n - 한국어-영어 Pararell Corpus를 활용한 한국어-영어 지식연결 (사전학습)\n - 한국어 문화, 언어를 고려해 언어학자가 제작한 데이터를 활용한 미세조정\n - 강화학습\n이 모든게 한꺼번에 적용되고 상업적 이용이 가능한 Bllossom을 이용해 여러분 만의 모델을 만들어보세욥!\nGPU가 부족하면 양자화 모델로 바로 서비스를 활용해 보세요 [양자화모델](https://huggingface.co/Bllossom/llama-3-Korean-Bllossom-70B-gguf-Q4_K_M)!!\n\n1. Bllossom-70.8B는 서울과기대, 테디썸, 연세대 언어자원 연구실의 언어학자와 협업해 만든 실용주의기반 언어모델입니다! 앞으로 지속적인 업데이트를 통해 관리하겠습니다 많이 활용해주세요 🙂\n2. 초 강력한 Advanced-Bllossom 8B, 70B모델, 시각-언어모델을 보유하고 있습니다! (궁금하신분은 개별 연락주세요!!)\n3. Bllossom은 NAACL2024, LREC-COLING2024 (구두) 발표로 채택되었습니다.\n4. 좋은 언어모델 계속 업데이트 하겠습니다!! 한국어 강화를위해 공동 연구하실분(특히논문) 언제든 환영합니다!! \n   특히 소량의 GPU라도 대여 가능한팀은 언제든 연락주세요! 만들고 싶은거 도와드려요.\n```\n\nThe Bllossom language model is a Korean-English bilingual language model based on the open-source LLama3. It enhances the connection of knowledge between Korean and English. It has the following features:\n\n* **Knowledge Linking**: Linking Korean and English knowledge through additional training\n* **Vocabulary Expansion**: Expansion of Korean vocabulary to enhance Korean expressiveness.\n* **Instruction Tuning**: Tuning using custom-made instruction following data specialized for Korean language and Korean culture\n* **Human Feedback**: DPO has been applied\n* **Vision-Language Alignment**: Aligning the vision transformer with this language model \n\n**This model developed by [MLPLab at Seoultech](http://mlp.seoultech.ac.kr), [Teddysum](http://teddysum.ai/) and [Yonsei Univ](https://sites.google.com/view/hansaemkim/hansaem-kim)**\n\n## Demo Video\n\n<div style=\"display: flex; justify-content: space-between;\">\n  <!-- 첫 번째 컬럼 -->\n  <div style=\"width: 49%;\">\n    <a>\n      <img src=\"https://github.com/lhsstn/lhsstn/blob/main/x-llava_dem.gif?raw=true\" style=\"width: 100%; height: auto;\">\n    </a>\n    <p style=\"text-align: center;\">Bllossom-V Demo</p>\n  </div>\n\n  <!-- 두 번째 컬럼 (필요하다면) -->\n  <div style=\"width: 49%;\">\n    <a>\n      <img src=\"https://github.com/lhsstn/lhsstn/blob/main/bllossom_demo_kakao.gif?raw=true\" style=\"width: 70%; height: auto;\">\n    </a>\n    <p style=\"text-align: center;\">Bllossom Demo(Kakao)ㅤㅤㅤㅤㅤㅤㅤㅤ</p>\n  </div>\n</div>\n\n\n## Example code\n\n### Colab Tutorial\n - [Inference-Code-Link](https://colab.research.google.com/drive/1fBOzUVZ6NRKk_ugeoTbAOokWKqSN47IG?usp=sharing)\n\n### Install Dependencies\n```bash\npip install torch transformers==4.40.0 accelerate\n```\n\n### Python code with Pipeline\n```python\nimport transformers\nimport torch\n\nmodel_id = \"Bllossom/llama-3-Korean-Bllossom-70B\"\n\npipeline = transformers.pipeline(\n    \"text-generation\",\n    model=model_id,\n    model_kwargs={\"torch_dtype\": torch.bfloat16},\n    device_map=\"auto\",\n)\n\npipeline.model.eval()\nPROMPT = '''You are a helpful AI assistant. Please answer the user's questions kindly. 당신은 유능한 AI 어시스턴트 입니다. 사용자의 질문에 대해 친절하게 답변해주세요.'''\ninstruction = \"서울과학기술대학교 MLP연구실에 대해 소개해줘\"\n\nmessages = [\n    {\"role\": \"system\", \"content\": f\"{PROMPT}\"},\n    {\"role\": \"user\", \"content\": f\"{instruction}\"}\n    ]\n\nprompt = pipeline.tokenizer.apply_chat_template(\n        messages, \n        tokenize=False, \n        add_generation_prompt=True\n)\n\nterminators = [\n    pipeline.tokenizer.eos_token_id,\n    pipeline.tokenizer.convert_tokens_to_ids(\"<|eot_id|>\")\n]\n\noutputs = pipeline(\n    prompt,\n    max_new_tokens=2048,\n    eos_token_id=terminators,\n    do_sample=True,\n    temperature=0.6,\n    top_p=0.9,\n)\n\nprint(outputs[0][\"generated_text\"][len(prompt):])\n\n# 서울과학기술대학교 MLP연구실은 멀티모달 자연어처리 연구를 하고 있습니다. 구성원은 임경태 교수와 김민준, 김상민, 최창수, 원인호, 유한결, 임현석, 송승우, 육정훈, 신동재 학생이 있습니다.\n```\n\n### Python code with AutoModel\n```python\n\nimport os\nimport torch\nfrom transformers import AutoTokenizer, AutoModelForCausalLM\n\nmodel_id = 'Bllossom/llama-3-Korean-Bllossom-70B'\n\ntokenizer = AutoTokenizer.from_pretrained(model_id)\nmodel = AutoModelForCausalLM.from_pretrained(\n    model_id,\n    torch_dtype=torch.bfloat16,\n    device_map=\"auto\",\n)\n\nmodel.eval()\n\nPROMPT = '''You are a helpful AI assistant. Please answer the user's questions kindly. 당신은 유능한 AI 어시스턴트 입니다. 사용자의 질문에 대해 친절하게 답변해주세요.'''\ninstruction = \"서울과학기술대학교 MLP연구실에 대해 소개해줘\"\n\nmessages = [\n    {\"role\": \"system\", \"content\": f\"{PROMPT}\"},\n    {\"role\": \"user\", \"content\": f\"{instruction}\"}\n    ]\n\ninput_ids = tokenizer.apply_chat_template(\n    messages,\n    add_generation_prompt=True,\n    return_tensors=\"pt\"\n).to(model.device)\n\nterminators = [\n    tokenizer.eos_token_id,\n    tokenizer.convert_tokens_to_ids(\"<|eot_id|>\")\n]\n\noutputs = model.generate(\n    input_ids,\n    max_new_tokens=2048,\n    eos_token_id=terminators,\n    do_sample=True,\n    temperature=0.6,\n    top_p=0.9\n)\n\nprint(tokenizer.decode(outputs[0][input_ids.shape[-1]:], skip_special_tokens=True))\n# 서울과학기술대학교 MLP연구실은 멀티모달 자연어처리 연구를 하고 있습니다. 구성원은 임경태 교수와 김민준, 김상민, 최창수, 원인호, 유한결, 임현석, 송승우, 육정훈, 신동재 학생이 있습니다.\n```\n\n\n\n## Citation\n**Language Model**\n```text\n@misc{bllossom,\n  author = {ChangSu Choi, Yongbin Jeong, Seoyoon Park, InHo Won, HyeonSeok Lim, SangMin Kim, Yejee Kang, Chanhyuk Yoon, Jaewan Park, Yiseul Lee, HyeJin Lee, Younggyun Hahm, Hansaem Kim, KyungTae Lim},\n  title = {Optimizing Language Augmentation for Multilingual Large Language Models: A Case Study on Korean},\n  year = {2024},\n  journal = {LREC-COLING 2024},\n  paperLink = {\\url{https://arxiv.org/pdf/2403.10882}},\n },\n}\n```\n\n**Vision-Language Model**\n```text\n@misc{bllossom-V,\n  author = {Dongjae Shin, Hyunseok Lim, Inho Won, Changsu Choi, Minjun Kim, Seungwoo Song, Hangyeol Yoo, Sangmin Kim, Kyungtae Lim},\n  title = {X-LLaVA: Optimizing Bilingual Large Vision-Language Alignment},\n  year = {2024},\n  publisher = {GitHub},\n  journal = {NAACL 2024 findings},\n  paperLink = {\\url{https://arxiv.org/pdf/2403.11399}},\n },\n}\n```\n\n## Contact\n - 임경태(KyungTae Lim), Professor at Seoultech. `ktlim@seoultech.ac.kr`\n - 함영균(Younggyun Hahm), CEO of Teddysum. `hahmyg@teddysum.ai`\n - 김한샘(Hansaem Kim), Professor at Yonsei. `khss@yonsei.ac.kr`\n\n## Contributor\n - 최창수(Chansu Choi), choics2623@seoultech.ac.kr\n - 김상민(Sangmin Kim), sangmin9708@naver.com\n - 원인호(Inho Won), wih1226@seoultech.ac.kr\n - 김민준(Minjun Kim), mjkmain@seoultech.ac.kr \n - 송승우(Seungwoo Song), sswoo@seoultech.ac.kr\n - 신동재(Dongjae Shin), dylan1998@seoultech.ac.kr\n - 임현석(Hyeonseok Lim), gustjrantk@seoultech.ac.kr\n - 육정훈(Jeonghun Yuk), usually670@gmail.com\n - 유한결(Hangyeol Yoo), 21102372@seoultech.ac.kr\n - 송서현(Seohyun Song), alexalex225225@gmail.com\n\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "arxiv:2403.10882",
    "arxiv:2403.11399",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 0,
  "downloads": 1133,
  "gated": false,
  "private": false,
  "last_modified": "2024-09-05T03:48:51.000Z",
  "created_at": "2024-09-04T07:47:45.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "66d810a1cd604edb69674074",
  "id": "RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf",
  "modelId": "RichardErkhov/Bllossom_-_llama-3-Korean-Bllossom-70B-gguf",
  "sha": "f4fe19942a504a4e506afc3d8ad0f485ee6cc6d4",
  "createdAt": "2024-09-04T07:47:45.000Z",
  "lastModified": "2024-09-05T03:48:51.000Z",
  "author": "RichardErkhov",
  "downloads": 1133,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 36
}