richarderkhov/tokyotech-llm_-_swallow-mx-8x7b-nve-v0.1-gguf overview

{
  "metadata": {},
  "card_data": {
    "frontmatter": {},
    "hero_image_url": "./logo.png",
    "summary": "Our Swallow-MX-8x7b-NVE-v0.1 model has undergone continuous pre-training from the Mixtral-8x7B-Instruct-v0.1, primarily with the addition of Japanese language data. !logo",
    "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\nSwallow-MX-8x7b-NVE-v0.1 - GGUF\n- Model creator: https://huggingface.co/tokyotech-llm/\n- Original model: https://huggingface.co/tokyotech-llm/Swallow-MX-8x7b-NVE-v0.1/\n\n\n| Name | Quant method | Size |\n| ---- | ---- | ---- |\n| [Swallow-MX-8x7b-NVE-v0.1.Q2_K.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q2_K.gguf) | Q2_K | 16.12GB |\n| [Swallow-MX-8x7b-NVE-v0.1.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.IQ3_XS.gguf) | IQ3_XS | 18.02GB |\n| [Swallow-MX-8x7b-NVE-v0.1.IQ3_S.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.IQ3_S.gguf) | IQ3_S | 19.03GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q3_K_S.gguf) | Q3_K_S | 19.03GB |\n| [Swallow-MX-8x7b-NVE-v0.1.IQ3_M.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.IQ3_M.gguf) | IQ3_M | 19.96GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q3_K.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q3_K.gguf) | Q3_K | 21.0GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q3_K_M.gguf) | Q3_K_M | 21.0GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q3_K_L.gguf) | Q3_K_L | 22.51GB |\n| [Swallow-MX-8x7b-NVE-v0.1.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.IQ4_XS.gguf) | IQ4_XS | 23.63GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q4_0.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q4_0.gguf) | Q4_0 | 24.63GB |\n| [Swallow-MX-8x7b-NVE-v0.1.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.IQ4_NL.gguf) | IQ4_NL | 24.91GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q4_K_S.gguf) | Q4_K_S | 24.91GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q4_K.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q4_K.gguf) | Q4_K | 26.49GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q4_K_M.gguf) | Q4_K_M | 26.49GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q4_1.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q4_1.gguf) | Q4_1 | 27.32GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q5_0.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q5_0.gguf) | Q5_0 | 30.02GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q5_K_S.gguf) | Q5_K_S | 30.02GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q5_K.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q5_K.gguf) | Q5_K | 30.95GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q5_K_M.gguf) | Q5_K_M | 30.95GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q5_1.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q5_1.gguf) | Q5_1 | 32.71GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q6_K.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/blob/main/Swallow-MX-8x7b-NVE-v0.1.Q6_K.gguf) | Q6_K | 35.74GB |\n| [Swallow-MX-8x7b-NVE-v0.1.Q8_0.gguf](https://huggingface.co/RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf/tree/main/) | Q8_0 | 46.22GB |\n\n\n\n\nOriginal model description:\n---\nlanguage:\n  - en\n  - ja\nlibrary_name: transformers\npipeline_tag: text-generation\ntag: moe\nlicense: apache-2.0\n---\n\n# Swallow-MX-8x7b-NVE-v0.1\n\nOur Swallow-MX-8x7b-NVE-v0.1 model has undergone continuous pre-training from the [Mixtral-8x7B-Instruct-v0.1](https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1), primarily with the addition of Japanese language data. \n\n![logo](./logo.png)\n\n## Model Details\n\n* **Model type**: Please refer to [Mixtral technical report](https://arxiv.org/abs/2401.04088) for details on the model architecture. \n* **Language(s)**: Japanese English\n* **Tokenizer**: This model utilizes the same tokenizer as employed by Mixtral-8x7B-Instruct-v0.1.\n* **Contact**: swallow[at]nlp.c.titech.ac.jp \n\n## Base Model Performance\n\n### Japanese version\n\n|Model|Size|JCommonsenseQA|JEMHopQA|NIILC|JSQuAD|XL-Sum|MGSM|WMT20-en-ja|WMT20-ja-en|\n|---|---|---|---|---|---|---|---|---|---|\n|   |   |4-shot|4-shot|4-shot|4-shot|1-shot|4-shot|4-shot|4-shot|\n| Llama 2 | 7B | 0.3852 | 0.4240 | 0.3410 | 0.7917 | 0.1905 | 0.0760 | 0.1783 | 0.1738 |\n| Swallow | 7B | 0.4808 | 0.5078 | 0.5968 | 0.8573 | 0.1830 | 0.1240 | 0.2510 | 0.1511 |\n| Swallow-Plus | 7B | 0.5478 | 0.5493 | 0.6030 | 0.8544 | 0.1806 | 0.1360 | 0.2568 | 0.1441 |\n| Swallow-NVE | 7B | 0.5433 | 0.5425 | 0.5729 | 0.8684 | 0.2117 | 0.1200 | 0.2405 | 0.1512 |\n| Mistral-7B-v0.1 |  7B | 0.7301 | 0.4245\t| 0.2722 | 0.8563 | 0.2006 | 0.1760 | 0.1405 | 0.1733 |\n|Swallow-MS-7b-v0.1| 7B | 0.8570 | 0.4915 | 0.5519 | 0.8802 | 0.1988 | 0.2240 | 0.2494 | 0.1667 |\n| Llama 2 | 13B | 0.6997 | 0.4415 | 0.4170 | 0.8533 | 0.2139 | 0.1320 | 0.2146 | 0.1982 |\n| Swallow | 13B | 0.7837 | 0.5063 | 0.6398 | 0.9005 | 0.2168 | 0.2040 | 0.2720 | 0.1771 |\n| Swallow-NVE | 13B | 0.7712 | 0.5438 | 0.6351 | 0.9030 | 0.2294 | 0.2120 | 0.2735 | 0.1817 |\n| Llama 2 | 70B | 0.8686 | 0.4656 | 0.5256 | 0.9080 | 0.2361 | 0.3560 | 0.2643 | **0.2398** |\n| Swallow | 70B | 0.9348 | **0.6290** | 0.6960 | 0.9176 | 0.2266 | **0.4840** | **0.3043** | 0.2298 |\n| Swallow-NVE | 70B | **0.9410** | 0.5759 | **0.7024** | **0.9254** | **0.2758** | 0.4720 | 0.3042 | 0.2322 |\n|Mixtral-8x7B-v0.1|8x7B|0.8347|0.5335|0.3549|0.8847|0.2192|0.3120|0.1970|0.1987|\n|Swallow-MX-8x7b-NVE-v0.1|8x7B|0.9258|0.5843|0.5687|0.9148|0.2589|0.4360|0.2705|0.2074|\n\n### English version\n\n|Model|Size|OpenBookQA|TriviaQA|HellaSwag|SQuAD2.0|XWINO|GSM8K|\n|---|---|---|---|---|---|---|---|\n|   |   |8-shot|8-shot|8-shot|8-shot|8-shot|8-shot|\n| Llama 2 | 7B    | 0.3580     | 0.6265   | 0.5860    | 0.3207   | 0.9049 | 0.1410 |\n| Swallow | 7B    | 0.3180     | 0.4836   | 0.5308    | 0.3125   | 0.8817 | 0.1130 |\n| Swallow-Plus | 7B | 0.3280     | 0.4558   | 0.5259    | 0.3134   | 0.8929 | 0.1061 |\n| Swallow-NVE | 7B | 0.3180     | 0.5079   | 0.5329    | 0.2919   | 0.8817 | 0.0986 |\n| Mistral-7B-v0.1 |  7B | 0.3660 | 0.7050 | 0.6264 | 0.3799 | 0.9157 | 0.3533 | 0.3440 | 0.5976 | 0.5810 | 0.3364 | 0.9037 | 0.2623 |\n|Swallow-MS-7b-v0.1| 7B | 0.3440 | 0.5976 | 0.5810 | 0.3364 | 0.9037 | 0.2623 |\n| Llama 2 | 13B   | 0.3760     | 0.7255   | 0.6148    | 0.3681   | 0.9140 | 0.2403 |\n| Swallow | 13B   | 0.3500     | 0.5852   | 0.5660    | 0.3406   | 0.9075 | 0.2039 |\n| Swallow-NVE | 13B | 0.3460     | 0.6025   | 0.5700    | 0.3478   | 0.9006 | 0.1751 |\n| Llama 2 | 70B   | **0.4280** | **0.8239** | **0.6742** | 0.3770 | **0.9290** | 0.5284 |\n| Swallow | 70B   | 0.4220     | 0.7756   | 0.6458    | 0.3745   | 0.9204 | 0.4867 |\n| Swallow-NVE | 70B | 0.4240     | 0.7817   | 0.6439    | 0.3451   | 0.9256 | 0.4943 |\n|Mixtral-8x7B-v0.1|8x7B|0.3960|0.7989|0.6678|**0.3842**|0.9204|**0.5747**|\n|Swallow-MX-8x7b-NVE-v0.1|8x7B|0.3740|0.7847|0.6520|0.3801|0.9170|0.5694|\n\nPlease note that Swallow-MX-8x7b-NVE-v0.1 is not derived from Mixtral-8x7B-v0.1, but rather underwent continued pre-training from Mixtral-8x7B-Instruct-v0.1.\n\n## Usage\n\nFirst install additional dependencies in [requirements.txt](./requirements.txt):\n\n```sh\npip install -r requirements.txt\n```\n\n### Use the base model\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\nimport torch\n\nmodel_name = \"tokyotech-llm/Swallow-MX-8x7b-NVE-v0.1\"\ntokenizer = AutoTokenizer.from_pretrained(model_name)\n\nmodel = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map=\"auto\")\nprompt = \"東京工業大学の主なキャンパスは、\"\ninput_ids = tokenizer.encode(\n    prompt,\n    add_special_tokens=False,\n    return_tensors=\"pt\"\n)\ntokens = model.generate(\n    input_ids.to(device=model.device),\n    max_new_tokens=128,\n    temperature=0.99,\n    top_p=0.95,\n    do_sample=True,\n)\n\nout = tokenizer.decode(tokens[0], skip_special_tokens=True)\nprint(out)\n```\n\n## Training Datasets\n\n### Continual Pre-Training\nThe following datasets were used for continual pre-training.\n\n- [Algebraic Stack](https://huggingface.co/datasets/EleutherAI/proof-pile-2)\n- [Japanese Wikipedia](https://dumps.wikimedia.org/other/cirrussearch)\n- [RefinedWeb](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)\n- [Swallow Corpus](https://arxiv.org/abs/2404.17733)\n- [The Pile](https://huggingface.co/datasets/EleutherAI/pile)\n- [The Vault](https://github.com/FSoft-AI4Code/TheVault)\n\n## Risks and Limitations\n\nThe models released here are still in the early stages of our research and development and have not been tuned to ensure outputs align with human intent and safety considerations.\n\n## Acknowledgements\n\nWe thank Mistral AI for releasing Mixtral-8x7B-Instruct-v0.1 under an open license for others to build on.\n\nOur project is supported by the [ABCI Large-scale Language Model Building Support Program](https://abci.ai/en/link/llm_support_program.html) of the National Institute of Advanced Industrial Science and Technology. \n\n## License\n\napache-2.0\n\n## Authors\n\nHere are the team members:\n- From [Okazaki Laboratory](https://www.nlp.c.titech.ac.jp/index.en.html), the following members:\n  - [Naoaki Okazaki](https://www.chokkan.org/index.ja.html)\n  - [Sakae Mizuki](https://s-mizuki-nlp.github.io/)\n  - [Hiroki Iida](https://meshidenn.github.io/)\n  - [Mengsay Loem](https://loem-ms.github.io/)\n  - [Shota Hirai](https://huggingface.co/Kotemo428)\n  - [Kakeru Hattori](https://aya-se.vercel.app/)\n  - [Masanari Ohi](https://twitter.com/stjohn2007)\n- From [YOKOTA Laboratory](https://www.rio.gsic.titech.ac.jp/en/index.html), the following members:\n  - [Rio Yokota](https://twitter.com/rioyokota)\n  - [Kazuki Fujii](https://twitter.com/okoge_kaz)\n  - [Taishi Nakamura](https://twitter.com/Setuna7777_2)\n\n## How to cite\n\nIf you find our work helpful, please feel free to cite us.\n\n```\n@inproceedings{Fujii:COLM2024,\n   title={Continual Pre-Training for Cross-Lingual LLM Adaptation:\nEnhancing Japanese Language Capabilities},\n   author={Kazuki Fujii and Taishi Nakamura and Mengsay Loem and Hiroki\nIida and Masanari Ohi and Kakeru Hattori and Hirai Shota and Sakae\nMizuki and Rio Yokota and Naoaki Okazaki},\n   booktitle=\"Proceedings of the First Conference on Language Modeling\",\n   series={COLM},\n   pages=\"(to appear)\",\n   year=\"2024\",\n   month=oct,\n   address={University of Pennsylvania, USA},\n}\n\n@inproceedings{Okazaki:COLM2024,\n   title={Building a Large Japanese Web Corpus for Large Language Models},\n   author={Naoaki Okazaki and Kakeru Hattori and Hirai Shota and Hiroki\nIida and Masanari Ohi and Kazuki Fujii and Taishi Nakamura and Mengsay\nLoem and Rio Yokota and Sakae Mizuki},\n   booktitle=\"Proceedings of the First Conference on Language Modeling\",\n   series={COLM},\n   pages=\"(to appear)\",\n   year=\"2024\",\n   month=oct,\n   address={University of Pennsylvania, USA},\n}\n```\n\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "arxiv:2401.04088",
    "arxiv:2404.17733",
    "endpoints_compatible",
    "region:us"
  ],
  "likes": 0,
  "downloads": 87,
  "gated": false,
  "private": false,
  "last_modified": "2024-10-08T15:30:40.000Z",
  "created_at": "2024-10-07T22:42:30.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "670463d61886407a72e96d1c",
  "id": "RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf",
  "modelId": "RichardErkhov/tokyotech-llm_-_Swallow-MX-8x7b-NVE-v0.1-gguf",
  "sha": "5e7c5637b0ec51db1b8c750a587cf4ae9ecb3c5b",
  "createdAt": "2024-10-07T22:42:30.000Z",
  "lastModified": "2024-10-08T15:30:40.000Z",
  "author": "RichardErkhov",
  "downloads": 87,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 25
}