richarderkhov/contextualai_-_archangel_sft_llama13b-gguf overview

{
  "metadata": {},
  "card_data": {
    "frontmatter": {},
    "hero_image_url": "https://gist.github.com/assets/29318529/fe2d8391-dbd1-4b7e-9dc4-7cb97e55bc06",
    "summary": "Quantization made by Richard Erkhov. Github Discord Request more models archangel_sft_llama13b - GGUF | Name | Quant method | Size | | ---- | ---- | ---- | | archangel_sft_llama13b.Q2_K.gguf | Q2_K | 4.52GB | | archangel_sft_llama13b.IQ3_XS.gguf | IQ3_XS | 4.99GB | | archangel_sft_llama13b.IQ3_S.gguf | IQ3_S | 5.27GB | | archangel_sft_llama13b.Q3_K_S.gguf | Q3_K_S | 5.27GB | | archangel_sft_llama13b.IQ3_M.gguf | IQ3_M | 5.57GB | | archangel_sft_llama13b.Q3_K.gguf | Q3_K | 5.9GB | | archangel_sft_llama13b.Q3_K_M.gguf | Q3_K_M | 5.9GB | | archangel_sft_llama13b.Q3_K_L.gguf | Q3_K_L | 6.45GB | | archangel_sft_llama13b.IQ4_XS.gguf | IQ4_XS | 6.54GB | | archangel_sft_llama13b.Q4_0.gguf | Q4_0 | 6.86GB | | archangel_sft_llama13b.IQ4_NL.gguf | IQ4_NL | 6.9GB | | archangel_sft_llama13b.Q4_K_S.gguf | Q4_K_S | 6.91GB | | archangel_sft_llama13b.Q4_K.gguf | Q4_K | 7.33GB | | archangel_sft_llama13b.Q4_K_M.gguf | Q4_K_M | 7.33GB | | archangel_sft_llama13b.Q4_1.gguf | Q4_1 | 7.61GB | | archangel_sft_llama13b.Q5_0.gguf | Q5_0 | 8.36GB | | archangel_sft_llama13b.Q5_K_S.gguf | Q5_K_S | 8.36GB | | archangel_sft_llama13b.Q5_K.gguf | Q5_K | 8.6GB | | archangel_sft_llama13b.Q5_K_M.gguf | Q5_K_M | 8.6GB | | archangel_sft_llama13b.Q5_1.gguf | Q5_1 | 9.1GB | | archangel_sft_llama13b.Q6_K.gguf | Q6_K | 9.95GB | | archangel_sft_llama13b.Q8_0.gguf | Q8_0 | 12.88GB | Original model description: --- license: apache-2.0 datasets: language: metrics: tags: --- !halos This repo contains the model checkpoints for: To prompt Archangel models, ensure that the format is consistent with that of TuluV2. For example, a prompt should be formatted as follows, where  corresponds to the human's role and  corresponds to the LLM's role. The human should speak first: ``  Hi! I'm looking for a cake recipe.  What kind of cake?  Chocolate cake.  ` Note that a beginning-of-sequence (BOS) token is automatically added by all Archangel models during tokenization and does not have to be added by you. No end-of-sequence (EOS) token is added to the prompt. For models trained with our conditional SFT model, the tokenizers have additional tokens  and  included in the embeddings. To generate with these control tokens in the context, postpend either to the prompt. Please refer to our code repository or blog which contains intructions for training your own HALOs and links to our model cards. If you find this repo or the technical paper useful in your research, please feel free to cite our work: ` @techreport{ethayarajh2023halos, author = {Ethayarajh, Kawin and Xu, Winnie, and Jurafsky, Dan and Kiela, Douwe}, title = {Human-Centered Loss Functions (HALOs)}, institution = {Contextual AI}, note = {https://github.com/ContextualAI/HALOs/blob/main/assets/report.pdf}, year = {2023}, } ``",
    "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\narchangel_sft_llama13b - GGUF\n- Model creator: https://huggingface.co/ContextualAI/\n- Original model: https://huggingface.co/ContextualAI/archangel_sft_llama13b/\n\n\n| Name | Quant method | Size |\n| ---- | ---- | ---- |\n| [archangel_sft_llama13b.Q2_K.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q2_K.gguf) | Q2_K | 4.52GB |\n| [archangel_sft_llama13b.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.IQ3_XS.gguf) | IQ3_XS | 4.99GB |\n| [archangel_sft_llama13b.IQ3_S.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.IQ3_S.gguf) | IQ3_S | 5.27GB |\n| [archangel_sft_llama13b.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q3_K_S.gguf) | Q3_K_S | 5.27GB |\n| [archangel_sft_llama13b.IQ3_M.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.IQ3_M.gguf) | IQ3_M | 5.57GB |\n| [archangel_sft_llama13b.Q3_K.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q3_K.gguf) | Q3_K | 5.9GB |\n| [archangel_sft_llama13b.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q3_K_M.gguf) | Q3_K_M | 5.9GB |\n| [archangel_sft_llama13b.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q3_K_L.gguf) | Q3_K_L | 6.45GB |\n| [archangel_sft_llama13b.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.IQ4_XS.gguf) | IQ4_XS | 6.54GB |\n| [archangel_sft_llama13b.Q4_0.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q4_0.gguf) | Q4_0 | 6.86GB |\n| [archangel_sft_llama13b.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.IQ4_NL.gguf) | IQ4_NL | 6.9GB |\n| [archangel_sft_llama13b.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q4_K_S.gguf) | Q4_K_S | 6.91GB |\n| [archangel_sft_llama13b.Q4_K.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q4_K.gguf) | Q4_K | 7.33GB |\n| [archangel_sft_llama13b.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q4_K_M.gguf) | Q4_K_M | 7.33GB |\n| [archangel_sft_llama13b.Q4_1.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q4_1.gguf) | Q4_1 | 7.61GB |\n| [archangel_sft_llama13b.Q5_0.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q5_0.gguf) | Q5_0 | 8.36GB |\n| [archangel_sft_llama13b.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q5_K_S.gguf) | Q5_K_S | 8.36GB |\n| [archangel_sft_llama13b.Q5_K.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q5_K.gguf) | Q5_K | 8.6GB |\n| [archangel_sft_llama13b.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q5_K_M.gguf) | Q5_K_M | 8.6GB |\n| [archangel_sft_llama13b.Q5_1.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q5_1.gguf) | Q5_1 | 9.1GB |\n| [archangel_sft_llama13b.Q6_K.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q6_K.gguf) | Q6_K | 9.95GB |\n| [archangel_sft_llama13b.Q8_0.gguf](https://huggingface.co/RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf/blob/main/archangel_sft_llama13b.Q8_0.gguf) | Q8_0 | 12.88GB |\n\n\n\n\nOriginal model description:\n---\nlicense: apache-2.0\ndatasets:\n- stanfordnlp/SHP\n- Anthropic/hh-rlhf\n- OpenAssistant/oasst1\nlanguage:\n- en\nmetrics:\n- accuracy\ntags:\n- human feedback\n- rlhf\n- preferences\n- alignment\n- HALO\n- halos\n- dpo\n- rl\n---\n\n![halos](https://gist.github.com/assets/29318529/fe2d8391-dbd1-4b7e-9dc4-7cb97e55bc06)\n\nThis repo contains the model checkpoints for:\n- model family <b>llama13b</b>\n- optimized with the loss <b>SFT</b>\n- aligned using the SHP, Anthropic HH and Open Assistant datasets.\n\nTo prompt Archangel models, ensure that the format is consistent with that of TuluV2.\nFor example, a prompt should be formatted as follows, where `<|user|>` corresponds to the human's role and `<|assistant|>` corresponds to the LLM's role.\nThe human should speak first:\n```\n<|user|>\nHi! I'm looking for a cake recipe.\n<|assistant|>\nWhat kind of cake?\n<|user|>\nChocolate cake.\n<|assistant|>\n```\nNote that a beginning-of-sequence (BOS) token is automatically added by all Archangel models during tokenization and does not have to be added by you. No end-of-sequence (EOS) token is added to the prompt.\n\nFor models trained with our conditional SFT model, the tokenizers have additional tokens `<|good|>` and `<|bad|>` included in the embeddings. \nTo generate with these control tokens in the context, postpend either to the prompt.\n\nPlease refer to our [code repository](https://github.com/ContextualAI/HALOs) or [blog](https://contextual.ai/better-cheaper-faster-llm-alignment-with-kto/) which contains intructions for training your own HALOs and links to our model cards.\n\nIf you find this repo or the technical paper useful in your research, please feel free to cite [our work](https://github.com/ContextualAI/HALOs/blob/main/assets/report.pdf):\n```\n@techreport{ethayarajh2023halos,\n  author = {Ethayarajh, Kawin and Xu, Winnie, and Jurafsky, Dan and Kiela, Douwe},\n  title = {Human-Centered Loss Functions (HALOs)},\n  institution = {Contextual AI},\n  note = {https://github.com/ContextualAI/HALOs/blob/main/assets/report.pdf},\n  year = {2023},\n}\n```\n\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "endpoints_compatible",
    "region:us"
  ],
  "likes": 0,
  "downloads": 160,
  "gated": false,
  "private": false,
  "last_modified": "2024-08-04T09:48:28.000Z",
  "created_at": "2024-08-04T03:22:26.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "66aef3f2f0bb52918914183c",
  "id": "RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf",
  "modelId": "RichardErkhov/ContextualAI_-_archangel_sft_llama13b-gguf",
  "sha": "966f9850346e9741e96bc65c0bdcebb90bfe7533",
  "createdAt": "2024-08-04T03:22:26.000Z",
  "lastModified": "2024-08-04T09:48:28.000Z",
  "author": "RichardErkhov",
  "downloads": 160,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 24
}