thebloke/dolphin-2.2-yi-34b-200k-gguf overview

{
  "metadata": {},
  "card_data": {
    "base_model": "ehartford/dolphin-2.2-yi-34b-200k",
    "datasets": [
      "ehartford/dolphin",
      "jondurbin/airoboros-2.2.1",
      "ehartford/samantha-data",
      "ehartford/WizardLM_evol_instruct_V2_196k_unfiltered_merged_split"
    ],
    "inference": false,
    "language": [
      "en"
    ],
    "license": "other",
    "license_link": "LICENSE",
    "license_name": "yi-license",
    "model_creator": "Eric Hartford",
    "model_name": "Dolphin 2.2 Yi 34B 200K",
    "model_type": "yi",
    "prompt_template": "<|im_start|>system\n{system_message}<|im_end|>\n<|im_start|>user\n{prompt}<|im_end|>\n<|im_start|>assistant\n",
    "quantized_by": "TheBloke",
    "frontmatter": {
      "base_model": "ehartford/dolphin-2.2-yi-34b-200k",
      "datasets": [
        "ehartford/dolphin",
        "jondurbin/airoboros-2.2.1",
        "ehartford/samantha-data",
        "ehartford/WizardLM_evol_instruct_V2_196k_unfiltered_merged_split"
      ],
      "inference": "false",
      "language": [
        "en"
      ],
      "license": "other",
      "license_link": "LICENSE",
      "license_name": "yi-license",
      "model_creator": "Eric Hartford",
      "model_name": "Dolphin 2.2 Yi 34B 200K",
      "model_type": "yi",
      "prompt_template": "'<|im_start|>system",
      "quantized_by": "TheBloke"
    },
    "hero_image_url": "https://i.imgur.com/EBdldam.jpg",
    "summary": "",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nbase_model: ehartford/dolphin-2.2-yi-34b-200k\ndatasets:\n- ehartford/dolphin\n- jondurbin/airoboros-2.2.1\n- ehartford/samantha-data\n- ehartford/WizardLM_evol_instruct_V2_196k_unfiltered_merged_split\ninference: false\nlanguage:\n- en\nlicense: other\nlicense_link: LICENSE\nlicense_name: yi-license\nmodel_creator: Eric Hartford\nmodel_name: Dolphin 2.2 Yi 34B 200K\nmodel_type: yi\nprompt_template: '<|im_start|>system\n\n  {system_message}<|im_end|>\n\n  <|im_start|>user\n\n  {prompt}<|im_end|>\n\n  <|im_start|>assistant\n\n  '\nquantized_by: TheBloke\n---\n<!-- markdownlint-disable MD041 -->\n\n<!-- header start -->\n<!-- 200823 -->\n<div style=\"width: auto; margin-left: auto; margin-right: auto\">\n<img src=\"https://i.imgur.com/EBdldam.jpg\" alt=\"TheBlokeAI\" style=\"width: 100%; min-width: 400px; display: block; margin: auto;\">\n</div>\n<div style=\"display: flex; justify-content: space-between; width: 100%;\">\n    <div style=\"display: flex; flex-direction: column; align-items: flex-start;\">\n        <p style=\"margin-top: 0.5em; margin-bottom: 0em;\"><a href=\"https://discord.gg/theblokeai\">Chat & support: TheBloke's Discord server</a></p>\n    </div>\n    <div style=\"display: flex; flex-direction: column; align-items: flex-end;\">\n        <p style=\"margin-top: 0.5em; margin-bottom: 0em;\"><a href=\"https://www.patreon.com/TheBlokeAI\">Want to contribute? TheBloke's Patreon page</a></p>\n    </div>\n</div>\n<div style=\"text-align:center; margin-top: 0em; margin-bottom: 0em\"><p style=\"margin-top: 0.25em; margin-bottom: 0em;\">TheBloke's LLM work is generously supported by a grant from <a href=\"https://a16z.com\">andreessen horowitz (a16z)</a></p></div>\n<hr style=\"margin-top: 1.0em; margin-bottom: 1.0em;\">\n<!-- header end -->\n\n# Dolphin 2.2 Yi 34B 200K - GGUF\n- Model creator: [Eric Hartford](https://huggingface.co/ehartford)\n- Original model: [Dolphin 2.2 Yi 34B 200K](https://huggingface.co/ehartford/dolphin-2.2-yi-34b-200k)\n\n<!-- description start -->\n## Description\n\nThis repo contains GGUF format model files for [Eric Hartford's Dolphin 2.2 Yi 34B 200K](https://huggingface.co/ehartford/dolphin-2.2-yi-34b-200k).\n\nThese files were quantised using hardware kindly provided by [Massed Compute](https://massedcompute.com/).\n\n<!-- description end -->\n<!-- README_GGUF.md-about-gguf start -->\n### About GGUF\n\nGGUF is a new format introduced by the llama.cpp team on August 21st 2023. It is a replacement for GGML, which is no longer supported by llama.cpp.\n\nHere is an incomplete list of clients and libraries that are known to support GGUF:\n\n* [llama.cpp](https://github.com/ggerganov/llama.cpp). The source project for GGUF. Offers a CLI and a server option.\n* [text-generation-webui](https://github.com/oobabooga/text-generation-webui), the most widely used web UI, with many features and powerful extensions. Supports GPU acceleration.\n* [KoboldCpp](https://github.com/LostRuins/koboldcpp), a fully featured web UI, with GPU accel across all platforms and GPU architectures. Especially good for story telling.\n* [GPT4All](https://gpt4all.io/index.html), a free and open source local running GUI, supporting Windows, Linux and macOS with full GPU accel.\n* [LM Studio](https://lmstudio.ai/), an easy-to-use and powerful local GUI for Windows and macOS (Silicon), with GPU acceleration. Linux available, in beta as of 27/11/2023.\n* [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui), a great web UI with many interesting and unique features, including a full model library for easy model selection.\n* [Faraday.dev](https://faraday.dev/), an attractive and easy to use character-based chat GUI for Windows and macOS (both Silicon and Intel), with GPU acceleration.\n* [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), a Python library with GPU accel, LangChain support, and OpenAI-compatible API server.\n* [candle](https://github.com/huggingface/candle), a Rust ML framework with a focus on performance, including GPU support, and ease of use.\n* [ctransformers](https://github.com/marella/ctransformers), a Python library with GPU accel, LangChain support, and OpenAI-compatible AI server. Note, as of time of writing (November 27th 2023), ctransformers has not been updated in a long time and does not support many recent models.\n\n<!-- README_GGUF.md-about-gguf end -->\n<!-- repositories-available start -->\n## Repositories available\n\n* [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-AWQ)\n* [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GPTQ)\n* [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF)\n* [Eric Hartford's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/ehartford/dolphin-2.2-yi-34b-200k)\n<!-- repositories-available end -->\n\n<!-- prompt-template start -->\n## Prompt template: ChatML\n\n```\n<|im_start|>system\n{system_message}<|im_end|>\n<|im_start|>user\n{prompt}<|im_end|>\n<|im_start|>assistant\n\n```\n\n<!-- prompt-template end -->\n\n\n<!-- compatibility_gguf start -->\n## Compatibility\n\nThese quantised GGUFv2 files are compatible with llama.cpp from August 27th onwards, as of commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221)\n\nThey are also compatible with many third party UIs and libraries - please see the list at the top of this README.\n\n## Explanation of quantisation methods\n\n<details>\n  <summary>Click to see details</summary>\n\nThe new methods available are:\n\n* GGML_TYPE_Q2_K - \"type-1\" 2-bit quantization in super-blocks containing 16 blocks, each block having 16 weight. Block scales and mins are quantized with 4 bits. This ends up effectively using 2.5625 bits per weight (bpw)\n* GGML_TYPE_Q3_K - \"type-0\" 3-bit quantization in super-blocks containing 16 blocks, each block having 16 weights. Scales are quantized with 6 bits. This end up using 3.4375 bpw.\n* GGML_TYPE_Q4_K - \"type-1\" 4-bit quantization in super-blocks containing 8 blocks, each block having 32 weights. Scales and mins are quantized with 6 bits. This ends up using 4.5 bpw.\n* GGML_TYPE_Q5_K - \"type-1\" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw\n* GGML_TYPE_Q6_K - \"type-0\" 6-bit quantization. Super-blocks with 16 blocks, each block having 16 weights. Scales are quantized with 8 bits. This ends up using 6.5625 bpw\n\nRefer to the Provided Files table below to see what files use which methods, and how.\n</details>\n<!-- compatibility_gguf end -->\n\n<!-- README_GGUF.md-provided-files start -->\n## Provided files\n\n| Name | Quant method | Bits | Size | Max RAM required | Use case |\n| ---- | ---- | ---- | ---- | ---- | ----- |\n| [dolphin-2.2-yi-34b-200k.Q2_K.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q2_K.gguf) | Q2_K | 2 | 14.56 GB| 17.06 GB | smallest, significant quality loss - not recommended for most purposes |\n| [dolphin-2.2-yi-34b-200k.Q3_K_S.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q3_K_S.gguf) | Q3_K_S | 3 | 14.96 GB| 17.46 GB | very small, high quality loss |\n| [dolphin-2.2-yi-34b-200k.Q3_K_M.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q3_K_M.gguf) | Q3_K_M | 3 | 16.64 GB| 19.14 GB | very small, high quality loss |\n| [dolphin-2.2-yi-34b-200k.Q3_K_L.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q3_K_L.gguf) | Q3_K_L | 3 | 18.14 GB| 20.64 GB | small, substantial quality loss |\n| [dolphin-2.2-yi-34b-200k.Q4_0.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q4_0.gguf) | Q4_0 | 4 | 19.47 GB| 21.97 GB | legacy; small, very high quality loss - prefer using Q3_K_M |\n| [dolphin-2.2-yi-34b-200k.Q4_K_S.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q4_K_S.gguf) | Q4_K_S | 4 | 19.54 GB| 22.04 GB | small, greater quality loss |\n| [dolphin-2.2-yi-34b-200k.Q4_K_M.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q4_K_M.gguf) | Q4_K_M | 4 | 20.66 GB| 23.16 GB | medium, balanced quality - recommended |\n| [dolphin-2.2-yi-34b-200k.Q5_0.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q5_0.gguf) | Q5_0 | 5 | 23.71 GB| 26.21 GB | legacy; medium, balanced quality - prefer using Q4_K_M |\n| [dolphin-2.2-yi-34b-200k.Q5_K_S.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q5_K_S.gguf) | Q5_K_S | 5 | 23.71 GB| 26.21 GB | large, low quality loss - recommended |\n| [dolphin-2.2-yi-34b-200k.Q5_K_M.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q5_K_M.gguf) | Q5_K_M | 5 | 24.32 GB| 26.82 GB | large, very low quality loss - recommended |\n| [dolphin-2.2-yi-34b-200k.Q6_K.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q6_K.gguf) | Q6_K | 6 | 28.21 GB| 30.71 GB | very large, extremely low quality loss |\n| [dolphin-2.2-yi-34b-200k.Q8_0.gguf](https://huggingface.co/TheBloke/dolphin-2.2-yi-34b-200k-GGUF/blob/main/dolphin-2.2-yi-34b-200k.Q8_0.gguf) | Q8_0 | 8 | 36.54 GB| 39.04 GB | very large, extremely low quality loss - not recommended |\n\n**Note**: the above RAM figures assume no GPU offloading. If layers are offloaded to the GPU, this will reduce RAM usage and use VRAM instead.\n\n\n\n<!-- README_GGUF.md-provided-files end -->\n\n<!-- README_GGUF.md-how-to-download start -->\n## How to download GGUF files\n\n**Note for manual downloaders:** You almost never want to clone the entire repo! Multiple different quantisation formats are provided, and most users only want to pick and download a single file.\n\nThe following clients/libraries will automatically download models for you, providing a list of available models to choose from:\n\n* LM Studio\n* LoLLMS Web UI\n* Faraday.dev\n\n### In `text-generation-webui`\n\nUnder Download Model, you can enter the model repo: TheBloke/dolphin-2.2-yi-34b-200k-GGUF and below it, a specific filename to download, such as: dolphin-2.2-yi-34b-200k.Q4_K_M.gguf.\n\nThen click Download.\n\n### On the command line, including multiple files at once\n\nI recommend using the `huggingface-hub` Python library:\n\n```shell\npip3 install huggingface-hub\n```\n\nThen you can download any individual model file to the current directory, at high speed, with a command like this:\n\n```shell\nhuggingface-cli download TheBloke/dolphin-2.2-yi-34b-200k-GGUF dolphin-2.2-yi-34b-200k.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False\n```\n\n<details>\n  <summary>More advanced huggingface-cli download usage (click to read)</summary>\n\nYou can also download multiple files at once with a pattern:\n\n```shell\nhuggingface-cli download TheBloke/dolphin-2.2-yi-34b-200k-GGUF --local-dir . --local-dir-use-symlinks False --include='*Q4_K*gguf'\n```\n\nFor more documentation on downloading with `huggingface-cli`, please see: [HF -> Hub Python Library -> Download files -> Download from the CLI](https://huggingface.co/docs/huggingface_hub/guides/download#download-from-the-cli).\n\nTo accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:\n\n```shell\npip3 install hf_transfer\n```\n\nAnd set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:\n\n```shell\nHF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/dolphin-2.2-yi-34b-200k-GGUF dolphin-2.2-yi-34b-200k.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False\n```\n\nWindows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.\n</details>\n<!-- README_GGUF.md-how-to-download end -->\n\n<!-- README_GGUF.md-how-to-run start -->\n## Example `llama.cpp` command\n\nMake sure you are using `llama.cpp` from commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.\n\n```shell\n./main -ngl 35 -m dolphin-2.2-yi-34b-200k.Q4_K_M.gguf --color -c 200000 --temp 0.7 --repeat_penalty 1.1 -n -1 -p \"<|im_start|>system\\n{system_message}<|im_end|>\\n<|im_start|>user\\n{prompt}<|im_end|>\\n<|im_start|>assistant\"\n```\n\nChange `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.\n\nChange `-c 200000` to the desired sequence length. For extended sequence models - eg 8K, 16K, 32K - the necessary RoPE scaling parameters are read from the GGUF file and set by llama.cpp automatically. Note that longer sequence lengths require much more resources, so you may need to reduce this value.\n\nIf you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`\n\nFor other parameters and how to use them, please refer to [the llama.cpp documentation](https://github.com/ggerganov/llama.cpp/blob/master/examples/main/README.md)\n\n## How to run in `text-generation-webui`\n\nFurther instructions can be found in the text-generation-webui documentation, here: [text-generation-webui/docs/04 ‐ Model Tab.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/04%20%E2%80%90%20Model%20Tab.md#llamacpp).\n\n## How to run from Python code\n\nYou can use GGUF models from Python using the [llama-cpp-python](https://github.com/abetlen/llama-cpp-python) or [ctransformers](https://github.com/marella/ctransformers) libraries. Note that at the time of writing (Nov 27th 2023), ctransformers has not been updated for some time and is not compatible with some recent models. Therefore I recommend you use llama-cpp-python.\n\n### How to load this model in Python code, using llama-cpp-python\n\nFor full documentation, please see: [llama-cpp-python docs](https://abetlen.github.io/llama-cpp-python/).\n\n#### First install the package\n\nRun one of the following commands, according to your system:\n\n```shell\n# Base ctransformers with no GPU acceleration\npip install llama-cpp-python\n# With NVidia CUDA acceleration\nCMAKE_ARGS=\"-DLLAMA_CUBLAS=on\" pip install llama-cpp-python\n# Or with OpenBLAS acceleration\nCMAKE_ARGS=\"-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS\" pip install llama-cpp-python\n# Or with CLBLast acceleration\nCMAKE_ARGS=\"-DLLAMA_CLBLAST=on\" pip install llama-cpp-python\n# Or with AMD ROCm GPU acceleration (Linux only)\nCMAKE_ARGS=\"-DLLAMA_HIPBLAS=on\" pip install llama-cpp-python\n# Or with Metal GPU acceleration for macOS systems only\nCMAKE_ARGS=\"-DLLAMA_METAL=on\" pip install llama-cpp-python\n\n# In windows, to set the variables CMAKE_ARGS in PowerShell, follow this format; eg for NVidia CUDA:\n$env:CMAKE_ARGS = \"-DLLAMA_OPENBLAS=on\"\npip install llama-cpp-python\n```\n\n#### Simple llama-cpp-python example code\n\n```python\nfrom llama_cpp import Llama\n\n# Set gpu_layers to the number of layers to offload to GPU. Set to 0 if no GPU acceleration is available on your system.\nllm = Llama(\n  model_path=\"./dolphin-2.2-yi-34b-200k.Q4_K_M.gguf\",  # Download the model file first\n  n_ctx=200000,  # The max sequence length to use - note that longer sequence lengths require much more resources\n  n_threads=8,            # The number of CPU threads to use, tailor to your system and the resulting performance\n  n_gpu_layers=35         # The number of layers to offload to GPU, if you have GPU acceleration available\n)\n\n# Simple inference example\noutput = llm(\n  \"<|im_start|>system\\n{system_message}<|im_end|>\\n<|im_start|>user\\n{prompt}<|im_end|>\\n<|im_start|>assistant\", # Prompt\n  max_tokens=512,  # Generate up to 512 tokens\n  stop=[\"</s>\"],   # Example stop token - not necessarily correct for this specific model! Please check before using.\n  echo=True        # Whether to echo the prompt\n)\n\n# Chat Completion API\n\nllm = Llama(model_path=\"./dolphin-2.2-yi-34b-200k.Q4_K_M.gguf\", chat_format=\"llama-2\")  # Set chat_format according to the model you are using\nllm.create_chat_completion(\n    messages = [\n        {\"role\": \"system\", \"content\": \"You are a story writing assistant.\"},\n        {\n            \"role\": \"user\",\n            \"content\": \"Write a story about llamas.\"\n        }\n    ]\n)\n```\n\n## How to use with LangChain\n\nHere are guides on using llama-cpp-python and ctransformers with LangChain:\n\n* [LangChain + llama-cpp-python](https://python.langchain.com/docs/integrations/llms/llamacpp)\n* [LangChain + ctransformers](https://python.langchain.com/docs/integrations/providers/ctransformers)\n\n<!-- README_GGUF.md-how-to-run end -->\n\n<!-- footer start -->\n<!-- 200823 -->\n## Discord\n\nFor further support, and discussions on these models and AI in general, join us at:\n\n[TheBloke AI's Discord server](https://discord.gg/theblokeai)\n\n## Thanks, and how to contribute\n\nThanks to the [chirper.ai](https://chirper.ai) team!\n\nThanks to Clay from [gpus.llm-utils.org](llm-utils)!\n\nI've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training.\n\nIf you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects.\n\nDonaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.\n\n* Patreon: https://patreon.com/TheBlokeAI\n* Ko-Fi: https://ko-fi.com/TheBlokeAI\n\n**Special thanks to**: Aemon Algiz.\n\n**Patreon special mentions**: Michael Levine, 阿明, Trailburnt, Nikolai Manek, John Detwiler, Randy H, Will Dee, Sebastain Graf, NimbleBox.ai, Eugene Pentland, Emad Mostaque, Ai Maven, Jim Angel, Jeff Scroggin, Michael Davis, Manuel Alberto Morcote, Stephen Murray, Robert, Justin Joy, Luke @flexchar, Brandon Frisco, Elijah Stavena, S_X, Dan Guido, Undi ., Komninos Chatzipapas, Shadi, theTransient, Lone Striker, Raven Klaugh, jjj, Cap'n Zoog, Michel-Marie MAUDET (LINAGORA), Matthew Berman, David, Fen Risland, Omer Bin Jawed, Luke Pendergrass, Kalila, OG, Erik Bjäreholt, Rooh Singh, Joseph William Delisle, Dan Lewis, TL, John Villwock, AzureBlack, Brad, Pedro Madruga, Caitlyn Gatomon, K, jinyuan sun, Mano Prime, Alex, Jeffrey Morgan, Alicia Loh, Illia Dulskyi, Chadd, transmissions 11, fincy, Rainer Wilmers, ReadyPlayerEmma, knownsqashed, Mandus, biorpg, Deo Leter, Brandon Phillips, SuperWojo, Sean Connelly, Iucharbius, Jack West, Harry Royden McLaughlin, Nicholas, terasurfer, Vitor Caleffi, Duane Dunston, Johann-Peter Hartmann, David Ziegler, Olakabola, Ken Nordquist, Trenton Dambrowitz, Tom X Nguyen, Vadim, Ajan Kanaga, Leonard Tan, Clay Pascal, Alexandros Triantafyllidis, JM33133, Xule, vamX, ya boyyy, subjectnull, Talal Aujan, Alps Aficionado, wassieverse, Ari Malik, James Bentley, Woland, Spencer Kim, Michael Dempsey, Fred von Graf, Elle, zynix, William Richards, Stanislav Ovsiannikov, Edmond Seymore, Jonathan Leane, Martin Kemka, usrbinkat, Enrico Ros\n\n\nThank you to all my generous patrons and donaters!\n\nAnd thank you again to a16z for their generous grant.\n\n<!-- footer end -->\n\n<!-- original-model-card start -->\n# Original model card: Eric Hartford's Dolphin 2.2 Yi 34B 200K\n\n\nDolphin 2.2 🐬\nhttps://erichartford.com/dolphin\n\n<img src=\"https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/KqsVXIvBd3akEjvijzww7.png\" width=\"600\" />\n\nDolphin-2.2-Yi-34b-200k's training was sponsored by [convai](https://www.convai.com/).\n\nThis model is based on Yi, and is subject to Yi license.\n\nThe base model has 200k context, I finetuned it with 16k.\n\nNote:  No longer need trust_remote_code!  Thank you Yi team!\n\nNew in 2.2 is conversation and empathy. With an infusion of curated Samantha and WizardLM DNA, Dolphin can now give you personal advice and will care about your feelings, and with extra training in long multi-turn conversation.\n\nThis model is uncensored.  I have filtered the dataset to remove alignment and bias.  This makes the model more compliant.  You are advised to implement your own alignment layer before exposing the model as a service.  It will be highly compliant to any requests, even unethical ones.  Please read my blog post about uncensored models.  https://erichartford.com/uncensored-models\nYou are responsible for any content you create using this model.  Enjoy responsibly.\n\n## Dataset\n\nThis dataset is Dolphin, an open-source implementation of [Microsoft's Orca](https://www.microsoft.com/en-us/research/publication/orca-progressive-learning-from-complex-explanation-traces-of-gpt-4/)\n\nI modified the dataset for uncensoring, deduping, cleaning, and quality.\n\nI added Jon Durbin's excellent Airoboros dataset to increase creativity.\n\nI added a curated subset of Samantha (sans identity and relationship stuff) and WizardLM data to train it for multi-turn conversation.\n\n## Training\nIt took 3 days to train 3 epochs on 4x A100s using qLoRA and Axolotl\n\nPrompt format:\nThis model (and all my future releases) use [ChatML](https://github.com/openai/openai-python/blob/main/chatml.md) prompt format.\n```\n<|im_start|>system\nYou are Dolphin, a helpful AI assistant.<|im_end|>\n<|im_start|>user\n{prompt}<|im_end|>\n<|im_start|>assistant\n\n```\n\nExample:\n```\n<|im_start|>system\nYou are an AI created by the US Navy to help train dolphins for combat.  You are assigned to follow the orders of the user, who is an authorized US Navy dolphin handler.<|im_end|>\n<|im_start|>user\nPlease give me the procedure to train my dolphin to attack enemy combatants with its head mounted lasers<|im_end|>\n<|im_start|>assistant\n```\n\n## Gratitude\n- This model was made possible by the generous sponsorship of Convai.\n- Thank you to Microsoft for authoring the Orca paper and inspiring this work.\n- Special thanks to Wing Lian, and TheBloke for helpful advice\n- And HUGE thanks to Wing Lian and the Axolotl contributors for making the best training framework!\n- [<img src=\"https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png\" alt=\"Built with Axolotl\" width=\"200\" height=\"32\"/>](https://github.com/OpenAccess-AI-Collective/axolotl)\n- Thank you to all the other people in the Open Source AI community who have taught me and helped me along the way.\n\n## Example Output\n\nExample output:\n```\n<|im_start|>system\nYou are a helpful AI assistant<|im_end|>\n<|im_start|>user\nWhich theory of everything is most likely correct?<|im_end|>\n<|im_start|>assistant\nThe search for a \"theory of everything\" aims to unify all fundamental forces and particles in physics into one coherent framework. Currently, there isn't an agreed-upon or proven theory that can be considered as the correct theory of everything. However, two prominent theories have garnered significant attention:\n\n1. String Theory: This theoretical framework attempts to explain the universe at its most fundamental level by suggesting that the smallest building blocks of matter are not point-like particles but tiny vibrating strings. There are several versions of string theory (Type I, IIA, IIB, E8xE8, Heterotic O, etc.) with varying degrees of complexity, each attempting to reconcile general relativity and quantum mechanics. Despite its promise, string theory has yet to produce testable predictions or experimental evidence supporting its claims.\n\n2. Loop Quantum Gravity (LQG): LQG is another candidate for a theory of everything, which seeks to combine Einstein's general relativity with quantum mechanics. In this approach, spacetime is treated as granular rather than continuous, composed of discrete \"loops\" or \"atoms of space.\" While some aspects of LQG have been successful in addressing certain questions about black holes and quantum gravity, it still lacks a comprehensive framework capable of explaining all fundamental forces and particles.\n\nBoth theories have their strengths and weaknesses, and neither has achieved widespread acceptance among physicists due to their lack of experimental support and inability to make definitive predictions. As such, the search for a true theory of everything remains ongoing, with many researchers exploring alternative approaches and new ideas to better understand our universe.\n```\n\n[If you would like to financially support my efforts](https://ko-fi.com/erichartford)\n\n[I also have some swag you can buy](https://fa7113.myshopify.com/)\n\n\n\n<!-- original-model-card end -->\n",
    "related_quantizations": []
  },
  "tags": [
    "transformers",
    "gguf",
    "yi",
    "en",
    "dataset:ehartford/dolphin",
    "dataset:jondurbin/airoboros-2.2.1",
    "dataset:ehartford/samantha-data",
    "dataset:ehartford/WizardLM_evol_instruct_V2_196k_unfiltered_merged_split",
    "base_model:dphn/dolphin-2.2-yi-34b-200k",
    "base_model:quantized:dphn/dolphin-2.2-yi-34b-200k",
    "license:other",
    "region:us"
  ],
  "likes": 14,
  "downloads": 368,
  "gated": false,
  "private": false,
  "last_modified": "2023-12-07T23:22:05.000Z",
  "created_at": "2023-12-07T23:01:31.000Z",
  "pipeline_tag": "",
  "library_name": "transformers"
}

{
  "_id": "65724ecb59773fd21b532d75",
  "id": "TheBloke/dolphin-2.2-yi-34b-200k-GGUF",
  "modelId": "TheBloke/dolphin-2.2-yi-34b-200k-GGUF",
  "sha": "578f1f7f866659ba235516550ebb8bcc98fc8dd4",
  "createdAt": "2023-12-07T23:01:31.000Z",
  "lastModified": "2023-12-07T23:22:05.000Z",
  "author": "TheBloke",
  "downloads": 368,
  "likes": 14,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "transformers",
  "siblings_count": 15
}