richarderkhov/sail_-_sailor-1.8b-gguf overview

{
  "metadata": {},
  "card_data": {
    "frontmatter": {},
    "hero_image_url": "banner_sailor.jpg",
    "summary": "Sailor is distributed under the terms of the Apache License 2.0. No restrict on the research and the commercial use, but should comply with the Qwen License.",
    "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\nSailor-1.8B - GGUF\n- Model creator: https://huggingface.co/sail/\n- Original model: https://huggingface.co/sail/Sailor-1.8B/\n\n\n| Name | Quant method | Size |\n| ---- | ---- | ---- |\n| [Sailor-1.8B.Q2_K.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q2_K.gguf) | Q2_K | 0.79GB |\n| [Sailor-1.8B.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.IQ3_XS.gguf) | IQ3_XS | 0.86GB |\n| [Sailor-1.8B.IQ3_S.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.IQ3_S.gguf) | IQ3_S | 0.89GB |\n| [Sailor-1.8B.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q3_K_S.gguf) | Q3_K_S | 0.89GB |\n| [Sailor-1.8B.IQ3_M.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.IQ3_M.gguf) | IQ3_M | 0.92GB |\n| [Sailor-1.8B.Q3_K.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q3_K.gguf) | Q3_K | 0.95GB |\n| [Sailor-1.8B.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q3_K_M.gguf) | Q3_K_M | 0.95GB |\n| [Sailor-1.8B.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q3_K_L.gguf) | Q3_K_L | 0.98GB |\n| [Sailor-1.8B.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.IQ4_XS.gguf) | IQ4_XS | 1.01GB |\n| [Sailor-1.8B.Q4_0.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q4_0.gguf) | Q4_0 | 1.04GB |\n| [Sailor-1.8B.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.IQ4_NL.gguf) | IQ4_NL | 1.05GB |\n| [Sailor-1.8B.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q4_K_S.gguf) | Q4_K_S | 1.08GB |\n| [Sailor-1.8B.Q4_K.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q4_K.gguf) | Q4_K | 1.13GB |\n| [Sailor-1.8B.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q4_K_M.gguf) | Q4_K_M | 1.13GB |\n| [Sailor-1.8B.Q4_1.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q4_1.gguf) | Q4_1 | 1.13GB |\n| [Sailor-1.8B.Q5_0.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q5_0.gguf) | Q5_0 | 1.22GB |\n| [Sailor-1.8B.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q5_K_S.gguf) | Q5_K_S | 1.24GB |\n| [Sailor-1.8B.Q5_K.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q5_K.gguf) | Q5_K | 1.28GB |\n| [Sailor-1.8B.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q5_K_M.gguf) | Q5_K_M | 1.28GB |\n| [Sailor-1.8B.Q5_1.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q5_1.gguf) | Q5_1 | 1.31GB |\n| [Sailor-1.8B.Q6_K.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q6_K.gguf) | Q6_K | 1.47GB |\n| [Sailor-1.8B.Q8_0.gguf](https://huggingface.co/RichardErkhov/sail_-_Sailor-1.8B-gguf/blob/main/Sailor-1.8B.Q8_0.gguf) | Q8_0 | 1.82GB |\n\n\n\n\nOriginal model description:\n---\nlanguage:\n- en\n- zh\n- id\n- th\n- vi\n- ms\n- lo\ndatasets:\n- cerebras/SlimPajama-627B\n- Skywork/SkyPile-150B\n- allenai/MADLAD-400\n- cc100\ntags:\n- multilingual\n- sea\n- sailor\nlicense: apache-2.0\nbase_model: Qwen/Qwen1.5-1.8B\ninference: false\nmodel-index:\n- name: Sailor-1.8B\n  results:\n  - task:\n      type: text-generation\n    dataset:\n      name: XQuAD-Thai\n      type: XQuAD-Thai\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 32.72\n    - name: F1 (3-Shot)\n      type: F1 (3-Shot)\n      value: 48.66\n  - task:\n      type: text-generation\n    dataset:\n      name: TyDiQA-Indonesian\n      type: TyDiQA-Indonesian\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 40.88\n    - name: F1 (3-Shot)\n      type: F1 (3-Shot)\n      value: 65.37\n  - task:\n      type: text-generation\n    dataset:\n      name: XQuAD-Vietnamese\n      type: XQuAD-Vietnamese\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 34.22\n    - name: F1 (3-Shot)\n      type: F1 (3-Shot)\n      value: 53.35\n  - task:\n      type: text-generation\n    dataset:\n      name: XCOPA-Thai\n      type: XCOPA-Thai\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 53.8\n  - task:\n      type: text-generation\n    dataset:\n      name: XCOPA-Indonesian\n      type: XCOPA-Indonesian\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 64.20\n  - task:\n      type: text-generation\n    dataset:\n      name: XCOPA-Vietnamese\n      type: XCOPA-Vietnamese\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 63.20\n  - task:\n      type: text-generation\n    dataset:\n      name: M3Exam-Thai\n      type: M3Exam-Thai\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 25.38\n  - task:\n      type: text-generation\n    dataset:\n      name: M3Exam-Indonesian\n      type: M3Exam-Indonesian\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 28.30\n  - task:\n      type: text-generation\n    dataset:\n      name: M3Exam-Vietnamese\n      type: M3Exam-Vietnamese\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 34.71\n  - task:\n      type: text-generation\n    dataset:\n      name: BELEBELE-Thai\n      type: BELEBELE-Thai\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 34.22\n  - task:\n      type: text-generation\n    dataset:\n      name: BELEBELE-Indonesian\n      type: BELEBELE-Indonesian\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 34.89\n  - task:\n      type: text-generation\n    dataset:\n      name: BELEBELE-Vietnamese\n      type: BELEBELE-Vietnamese\n    metrics:\n    - name: EM (3-Shot)\n      type: EM (3-Shot)\n      value: 35.33\n---\n\n<div align=\"center\">\n  <img src=\"banner_sailor.jpg\" width=\"700\"/>\n</div>\n\nSailor is a suite of Open Language Models tailored for South-East Asia (SEA), focusing on languages such as 🇮🇩Indonesian, 🇹🇭Thai, 🇻🇳Vietnamese, 🇲🇾Malay, and 🇱🇦Lao. \nDeveloped with careful data curation, Sailor models are designed to understand and generate text across diverse linguistic landscapes of SEA region. \nBuilt from [Qwen 1.5](https://huggingface.co/collections/Qwen/qwen15-65c0a2f577b1ecb76d786524) , Sailor encompasses models of varying sizes, spanning from 0.5B to 7B versions for different requirements. \nWe further fine-tune the base model with open-source datasets to get instruction-tuned models, namedly Sailor-Chat. \nBenchmarking results demonstrate Sailor's proficiency in tasks such as question answering, commonsense reasoning, and other tasks in SEA languages.\n\n> The logo was generated by MidJourney\n\n## Model Summary\n- **Model Collections:** [Base Model & Chat Model](https://huggingface.co/collections/sail/sailor-65e19a749f978976f1959825)\n- **Project Website:** [sailorllm.github.io](https://sailorllm.github.io/)\n- **Codebase:** [github.com/sail-sg/sailor-llm](https://github.com/sail-sg/sailor-llm)\n- **Technical Report:** [arxiv.org/pdf/2404.03608.pdf](https://arxiv.org/pdf/2404.03608.pdf) \n\n\n## Training details\nSailor is crafted by continually pre-training from language models like the remarkable Qwen 1.5 models, which already has a great performance on SEA languages. \nThe pre-training corpus heavily leverages the publicly available corpus, including \n[SlimPajama](https://huggingface.co/datasets/cerebras/SlimPajama-627B), \n[SkyPile](https://huggingface.co/datasets/Skywork/SkyPile-150B), \n[CC100](https://huggingface.co/datasets/cc100) and [MADLAD-400](https://huggingface.co/datasets/allenai/MADLAD-400).\n\nBy employing aggressive data deduplication and careful data cleaning on the collected corpus, we have attained a high-quality dataset spanning various languages. \nThrough systematic experiments to determine the weights of different languages, Sailor models undergo training from 200B to 400B tokens, tailored to different model sizes. \nThe approach boosts their performance on SEA languages while maintaining proficiency in English and Chinese without significant compromise. \nFinally, we continually pre-train the Qwen1.5-0.5B model with 400 Billion tokens, and other models with 200 Billion tokens to obtain the Sailor models.\n\n## Requirements\nThe code of Sailor has been in the latest Hugging face transformers and we advise you to install `transformers>=4.37.0`.\n\n## Quickstart\n\nHere provides a code snippet to show you how to load the tokenizer and model and how to generate contents.\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\ndevice = \"cuda\" # the device to load the model\n\nmodel = AutoModelForCausalLM.from_pretrained(\"sail/Sailor-1.8B\", device_map=\"auto\")\ntokenizer = AutoTokenizer.from_pretrained(\"sail/Sailor-1.8B\")\n\ninput_message = \"Model bahasa adalah model probabilistik\" \n### The given Indonesian input translates to 'A language model is a probabilistic model of.'\n\nmodel_inputs = tokenizer([input_message], return_tensors=\"pt\").to(device)\n\ngenerated_ids = model.generate(\n    model_inputs.input_ids,\n    max_new_tokens=64\n)\n\ngenerated_ids = [\n    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)\n]\n\nresponse = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\nprint(response)\n```\n\n# License\n\nSailor is distributed under the terms of the Apache License 2.0. \nNo restrict on the research and the commercial use, but should comply with the [Qwen License](https://huggingface.co/Qwen/Qwen1.5-1.8B/blob/main/LICENSE).\n\n## Citation\n\nIf you find sailor useful, please cite our work as follows:\n\n```\n@misc{dou2024sailor,\n      title={Sailor: Open Language Models for South-East Asia}, \n      author={Longxu Dou and Qian Liu and Guangtao Zeng and Jia Guo and Jiahui Zhou and Wei Lu and Min Lin},\n      year={2024},\n      eprint={2404.03608},\n      archivePrefix={arXiv},\n      primaryClass={cs.CL}\n}\n```\n\n# Contact Us\n\nIf you have any questions, please raise an issue or contact us at [doulx@sea.com](mailto:doulx@sea.com) or [liuqian@sea.com](mailto:liuqian@sea.com).\n\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "arxiv:2404.03608",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 0,
  "downloads": 535,
  "gated": false,
  "private": false,
  "last_modified": "2024-05-17T12:11:35.000Z",
  "created_at": "2024-05-17T11:48:04.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "664743f48a26d54f96e35eba",
  "id": "RichardErkhov/sail_-_Sailor-1.8B-gguf",
  "modelId": "RichardErkhov/sail_-_Sailor-1.8B-gguf",
  "sha": "328ca6802253fdc1e4b312471b89443e316ca8b9",
  "createdAt": "2024-05-17T11:48:04.000Z",
  "lastModified": "2024-05-17T12:11:35.000Z",
  "author": "RichardErkhov",
  "downloads": 535,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 24
}