TheBloke/sqlcoder-GGUF

1---2language:3- en4license: other5library_name: transformers6tags:7- code8metrics:9- code_eval10model_name: Sqlcoder11base_model: defog/sqlcoder12inference: false13model_creator: Defog.ai14model_type: starcoder15pipeline_tag: text-generation16prompt_template: '{prompt}17 18  '19quantized_by: TheBloke20---21 22<!-- header start -->23<!-- 200823 -->24<div style="width: auto; margin-left: auto; margin-right: auto">25<img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">26</div>27<div style="display: flex; justify-content: space-between; width: 100%;">28    <div style="display: flex; flex-direction: column; align-items: flex-start;">29        <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>30    </div>31    <div style="display: flex; flex-direction: column; align-items: flex-end;">32        <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>33    </div>34</div>35<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>36<hr style="margin-top: 1.0em; margin-bottom: 1.0em;">37<!-- header end -->38 39# Sqlcoder - GGUF40- Model creator: [Defog.ai](https://huggingface.co/defog)41- Original model: [Sqlcoder](https://huggingface.co/defog/sqlcoder)42 43<!-- description start -->44## Description45 46This repo contains GGUF format model files for [Defog.ai's Sqlcoder](https://huggingface.co/defog/sqlcoder).47 48<!-- description end -->49<!-- README_GGUF.md-about-gguf start -->50### About GGUF51 52GGUF 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.53 54Here is an incomplate list of clients and libraries that are known to support GGUF:55 56* [llama.cpp](https://github.com/ggerganov/llama.cpp). The source project for GGUF. Offers a CLI and a server option.57* [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.58* [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.59* [LM Studio](https://lmstudio.ai/), an easy-to-use and powerful local GUI for Windows and macOS (Silicon), with GPU acceleration.60* [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.61* [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.62* [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), a Python library with GPU accel, LangChain support, and OpenAI-compatible API server.63* [candle](https://github.com/huggingface/candle), a Rust ML framework with a focus on performance, including GPU support, and ease of use.64 65<!-- README_GGUF.md-about-gguf end -->66<!-- repositories-available start -->67## Repositories available68 69* [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/sqlcoder-GPTQ)70* [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/sqlcoder-GGUF)71* [Defog.ai's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/defog/sqlcoder)72<!-- repositories-available end -->73 74<!-- prompt-template start -->75## Prompt template: Unknown76 77```78{prompt}79 80```81 82<!-- prompt-template end -->83 84 85<!-- compatibility_gguf start -->86## Compatibility87 88These quantised GGUFv2 files are compatible with llama.cpp from August 27th onwards, as of commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221)89 90They are also compatible with many third party UIs and libraries - please see the list at the top of this README.91 92## Explanation of quantisation methods93<details>94  <summary>Click to see details</summary>95 96The new methods available are:97* 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)98* 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.99* 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.100* GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw101* 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 bpw102 103Refer to the Provided Files table below to see what files use which methods, and how.104</details>105<!-- compatibility_gguf end -->106 107<!-- README_GGUF.md-provided-files start -->108## Provided files109 110| Name | Quant method | Bits | Size | Max RAM required | Use case |111| ---- | ---- | ---- | ---- | ---- | ----- |112| [sqlcoder.Q2_K.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q2_K.gguf) | Q2_K | 2 | 6.73 GB| 9.23 GB | smallest, significant quality loss - not recommended for most purposes |113| [sqlcoder.Q3_K_S.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q3_K_S.gguf) | Q3_K_S | 3 | 6.93 GB| 9.43 GB | very small, high quality loss |114| [sqlcoder.Q3_K_M.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q3_K_M.gguf) | Q3_K_M | 3 | 8.18 GB| 10.68 GB | very small, high quality loss |115| [sqlcoder.Q4_0.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q4_0.gguf) | Q4_0 | 4 | 8.99 GB| 11.49 GB | legacy; small, very high quality loss - prefer using Q3_K_M |116| [sqlcoder.Q4_K_S.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q4_K_S.gguf) | Q4_K_S | 4 | 9.06 GB| 11.56 GB | small, greater quality loss |117| [sqlcoder.Q3_K_L.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q3_K_L.gguf) | Q3_K_L | 3 | 9.08 GB| 11.58 GB | small, substantial quality loss |118| [sqlcoder.Q4_K_M.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q4_K_M.gguf) | Q4_K_M | 4 | 9.96 GB| 12.46 GB | medium, balanced quality - recommended |119| [sqlcoder.Q5_0.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q5_0.gguf) | Q5_0 | 5 | 10.93 GB| 13.43 GB | legacy; medium, balanced quality - prefer using Q4_K_M |120| [sqlcoder.Q5_K_S.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q5_K_S.gguf) | Q5_K_S | 5 | 10.93 GB| 13.43 GB | large, low quality loss - recommended |121| [sqlcoder.Q5_K_M.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q5_K_M.gguf) | Q5_K_M | 5 | 11.54 GB| 14.04 GB | large, very low quality loss - recommended |122| [sqlcoder.Q6_K.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q6_K.gguf) | Q6_K | 6 | 12.99 GB| 15.49 GB | very large, extremely low quality loss |123| [sqlcoder.Q8_0.gguf](https://huggingface.co/TheBloke/sqlcoder-GGUF/blob/main/sqlcoder.Q8_0.gguf) | Q8_0 | 8 | 16.82 GB| 19.32 GB | very large, extremely low quality loss - not recommended |124 125**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.126 127 128 129<!-- README_GGUF.md-provided-files end -->130 131<!-- README_GGUF.md-how-to-download start -->132## How to download GGUF files133 134**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.135 136The following clients/libraries will automatically download models for you, providing a list of available models to choose from:137- LM Studio138- LoLLMS Web UI139- Faraday.dev140 141### In `text-generation-webui`142 143Under Download Model, you can enter the model repo: TheBloke/sqlcoder-GGUF and below it, a specific filename to download, such as: sqlcoder.Q4_K_M.gguf.144 145Then click Download.146 147### On the command line, including multiple files at once148 149I recommend using the `huggingface-hub` Python library:150 151```shell152pip3 install huggingface-hub153```154 155Then you can download any individual model file to the current directory, at high speed, with a command like this:156 157```shell158huggingface-cli download TheBloke/sqlcoder-GGUF sqlcoder.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False159```160 161<details>162  <summary>More advanced huggingface-cli download usage</summary>163 164You can also download multiple files at once with a pattern:165 166```shell167huggingface-cli download TheBloke/sqlcoder-GGUF --local-dir . --local-dir-use-symlinks False --include='*Q4_K*gguf'168```169 170For 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).171 172To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:173 174```shell175pip3 install hf_transfer176```177 178And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:179 180```shell181HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/sqlcoder-GGUF sqlcoder.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False182```183 184Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.185</details>186<!-- README_GGUF.md-how-to-download end -->187 188<!-- README_GGUF.md-how-to-run start -->189## Example `llama.cpp` command190 191Make sure you are using `llama.cpp` from commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.192 193```shell194./main -ngl 32 -m sqlcoder.Q4_K_M.gguf --color -c 2048 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "{prompt}"195```196 197Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.198 199Change `-c 2048` 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.200 201If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`202 203For 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)204 205## How to run in `text-generation-webui`206 207Further instructions here: [text-generation-webui/docs/llama.cpp.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/llama.cpp.md).208 209## How to use with LangChain210 211Here are guides on using llama-cpp-python and ctransformers with LangChain:212 213* [LangChain + llama-cpp-python](https://python.langchain.com/docs/integrations/llms/llamacpp)214 215<!-- README_GGUF.md-how-to-run end -->216 217<!-- footer start -->218<!-- 200823 -->219## Discord220 221For further support, and discussions on these models and AI in general, join us at:222 223[TheBloke AI's Discord server](https://discord.gg/theblokeai)224 225## Thanks, and how to contribute226 227Thanks to the [chirper.ai](https://chirper.ai) team!228 229Thanks to Clay from [gpus.llm-utils.org](llm-utils)!230 231I'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.232 233If 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.234 235Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.236 237* Patreon: https://patreon.com/TheBlokeAI238* Ko-Fi: https://ko-fi.com/TheBlokeAI239 240**Special thanks to**: Aemon Algiz.241 242**Patreon special mentions**: Alicia Loh, Stephen Murray, K, Ajan Kanaga, RoA, Magnesian, Deo Leter, Olakabola, Eugene Pentland, zynix, Deep Realms, Raymond Fosdick, Elijah Stavena, Iucharbius, Erik Bjäreholt, Luis Javier Navarrete Lozano, Nicholas, theTransient, John Detwiler, alfie_i, knownsqashed, Mano Prime, Willem Michiel, Enrico Ros, LangChain4j, OG, Michael Dempsey, Pierre Kircher, Pedro Madruga, James Bentley, Thomas Belote, Luke @flexchar, Leonard Tan, Johann-Peter Hartmann, Illia Dulskyi, Fen Risland, Chadd, S_X, Jeff Scroggin, Ken Nordquist, Sean Connelly, Artur Olbinski, Swaroop Kallakuri, Jack West, Ai Maven, David Ziegler, Russ Johnson, transmissions 11, John Villwock, Alps Aficionado, Clay Pascal, Viktor Bowallius, Subspace Studios, Rainer Wilmers, Trenton Dambrowitz, vamX, Michael Levine, 준교 김, Brandon Frisco, Kalila, Trailburnt, Randy H, Talal Aujan, Nathan Dryer, Vadim, 阿明, ReadyPlayerEmma, Tiffany J. Kim, George Stoitzev, Spencer Kim, Jerry Meng, Gabriel Tamborski, Cory Kujawski, Jeffrey Morgan, Spiking Neurons AB, Edmond Seymore, Alexandros Triantafyllidis, Lone Striker, Cap'n Zoog, Nikolai Manek, danny, ya boyyy, Derek Yates, usrbinkat, Mandus, TL, Nathan LeClaire, subjectnull, Imad Khwaja, webtim, Raven Klaugh, Asp the Wyvern, Gabriel Puliatti, Caitlyn Gatomon, Joseph William Delisle, Jonathan Leane, Luke Pendergrass, SuperWojo, Sebastain Graf, Will Dee, Fred von Graf, Andrey, Dan Guido, Daniel P. Andersen, Nitin Borwankar, Elle, Vitor Caleffi, biorpg, jjj, NimbleBox.ai, Pieter, Matthew Berman, terasurfer, Michael Davis, Alex, Stanislav Ovsiannikov243 244 245Thank you to all my generous patrons and donaters!246 247And thank you again to a16z for their generous grant.248 249<!-- footer end -->250 251<!-- original-model-card start -->252# Original model card: Defog.ai's Sqlcoder253 254 255# Defog SQLCoder256Defog's SQLCoder is a state-of-the-art LLM for converting natural language questions to SQL queries.257 258[Interactive Demo](https://defog.ai/sqlcoder-demo) | [♾️ Colab](https://colab.research.google.com/drive/1z4rmOEiFkxkMiecAWeTUlPl0OmKgfEu7) | [🐦 Twitter](https://twitter.com/defogdata)259 260## TL;DR261SQLCoder is a 15B parameter model that slightly outperforms `gpt-3.5-turbo` for natural language to SQL generation tasks on our [sql-eval](https://github.com/defog-ai/sql-eval) framework, and significantly outperforms all popular open-source models. It also significantly outperforms `text-davinci-003`, a model that's more than 10 times its size.262 263SQLCoder is fine-tuned on a base StarCoder model.264 265## Results on novel datasets not seen in training266| model   | perc_correct |267|-|-|268| gpt-4            | 74.3 |269| defog-sqlcoder   | 64.6 |270| gpt-3.5-turbo    | 60.6 |271| defog-easysql    | 57.1 |272| text-davinci-003 | 54.3 |273| wizardcoder      | 52.0 |274| starcoder        | 45.1 |275 276## License277The model weights have a `CC BY-SA 4.0` license, with OpenRAIL-M clauses for responsible use attached. The TL;DR is that you can use and modify the model for any purpose – including commercial use. However, if you modify the weights (for example, by fine-tuning), you must open-source your modified weights under the same `CC BY-SA 4.0` license terms.278 279## Training280Defog was trained on 10,537 human-curated questions across 2 epochs. These questions were based on 10 different schemas. None of the schemas in the training data were included in our evaluation framework.281 282Training happened in 2 phases. The first phase was on questions that were classified as "easy" or "medium" difficulty, and the second phase was on questions that were classified as "hard" or "extra hard" difficulty.283 284The results of training on our easy+medium data were stored in a model called `defog-easy`. We found that the additional training on hard+extra-hard data led to a 7 percentage point increase in performance.285 286## Results by question category287We classified each generated question into one of 5 categories. The table displays the percentage of questions answered correctly by each model, broken down by category.288| query_category | gpt-4 | defog-sqlcoder | gpt-3.5-turbo | defog-easy | text-davinci-003 | wizard-coder | star-coder |289|-|-|-|-|-|-|-|-|290| group_by | 82.9 | 77.1 | 71.4 | 62.9 | 62.9 | 68.6 | 54.3 |291| order_by | 71.4 | 65.7 | 60.0 | 68.6 | 60.0 | 54.3 | 57.1 |292| ratio | 62.9 | 57.1 | 48.6 | 40.0 | 37.1 | 22.9 | 17.1 |293| table_join | 74.3 | 57.1 | 60.0 | 54.3 | 51.4 | 54.3 | 51.4 |294| where | 80.0 | 65.7 | 62.9 | 60.0 | 60.0 | 60.0 | 45.7 |295 296## Using SQLCoder297You can use SQLCoder via the `transformers` library by downloading our model weights from the HuggingFace repo. We have added sample code for inference [here](./inference.py). You can also use a demo on our website [here](https://defog.ai/sqlcoder-demo), or run SQLCoder in Colab [here](https://colab.research.google.com/drive/13BIKsqHnPOBcQ-ba2p77L5saiepTIwu0#scrollTo=ZpbVgVHMkJvC)298 299## Hardware Requirements300SQLCoder has been tested on an A100 40GB GPU with `bfloat16` weights. You can also load an 8-bit quantized version of the model on consumer GPUs with 20GB or more of memory – like RTX 4090, RTX 3090, and Apple M2 Pro, M2 Max, or M2 Ultra Chips with 20GB or more of memory.301 302## Todo303 304- [x] Open-source the v1 model weights305- [ ] Train the model on more data, with higher data variance306- [ ] Tune the model further with Reward Modelling and RLHF307- [ ] Pretrain a model from scratch that specializes in SQL analysis308 309<!-- original-model-card end -->310