Initial GGML model commit

README.md

@@ -17,51 +17,84 @@ license: other
 </div>
 <!-- header end -->
 
-# Falcon 40B-Instruct GGML GGML
+# Falcon 40B-Instruct GGML  GGML
 
-These files are **experimental** GGML format model files for [Falcon 40B-Instruct GGML](https://huggingface.co/tiiuae/falcon-40b-instruct).
+These files are GGML format model files for [Falcon 40B-Instruct GGML ](https://huggingface.co/tiiuae/falcon-40b-instruct).
 
-These GGML files will **not** work in llama.cpp, and at the time of writing they will not work with any UI or library. They cannot be used from Python code.
-
-They currently only work using the basic command line test tool, compiled from Jan Ploski's fork of the ggml repo, where support for Falcon 40B GGML was first completed.
-
-They are therefore uploaded purely for initial evaluation and experimentation. Support for these GGMLs should improve in the near future.
+GGML files are for CPU + GPU inference using [llama.cpp](https://github.com/ggerganov/llama.cpp) and libraries and UIs which support this format, such as:
+* [text-generation-webui](https://github.com/oobabooga/text-generation-webui)
+* [KoboldCpp](https://github.com/LostRuins/koboldcpp)
+* [ParisNeo/GPT4All-UI](https://github.com/ParisNeo/gpt4all-ui)
+* [llama-cpp-python](https://github.com/abetlen/llama-cpp-python)
+* [ctransformers](https://github.com/marella/ctransformers)
 
 ## Repositories available
 
-* [4-bit GPTQ model for GPU inference](https://huggingface.co/TheBloke/falcon-40b-instruct-GPTQ)
-* [3-bit GPTQ model for GPU inference](https://huggingface.co/TheBloke/falcon-40b-instruct-3bit-GPTQ)
-* [2, 3, 4, 5, 6, 8-bit GGML models for CPU+GPU inference](https://huggingface.co/TheBloke/falcon-40b-instruct-GGML)
+* [4-bit GPTQ models for GPU inference](https://huggingface.co/TheBloke/falcon-40b-instruct-GPTQ)
+* [2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference](https://huggingface.co/TheBloke/falcon-40b-instruct-GGML)
 * [Unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/tiiuae/falcon-40b-instruct)
 
 <!-- compatibility_ggml start -->
 ## Compatibility
 
-To build the CLI tool necessary to use these GGML files, please follow the following steps:
+### Original llama.cpp quant methods: `q4_0, q4_1, q5_0, q5_1, q8_0`
 
-```
-git clone https://github.com/jploski/ggml falcon-ggml
-cd falcon-ggml
-git checkout falcon40b
-mkdir build && cd build && cmake .. && cmake --build . --config Release
-```
+I have quantized these 'original' quantisation methods using an older version of llama.cpp so that they remain compatible with llama.cpp as of May 19th, commit `2d5db48`.
 
-Then run a command like the following, adjusting params as required:
-```
-bin/falcon -m /path/to/Falcon-40b-Instruct.ggmlv3.q4_0.bin -t 10 -n 200 -p "write a story about llamas"
-```
+These are guaranteed to be compatbile with any UIs, tools and libraries released since late May.
+
+### New k-quant methods: `q2_K, q3_K_S, q3_K_M, q3_K_L, q4_K_S, q4_K_M, q5_K_S, q6_K`
+
+These new quantisation methods are compatible with llama.cpp as of June 6th, commit `2d43387`.
+
+They are now also compatible with recent releases of text-generation-webui, KoboldCpp, llama-cpp-python and ctransformers. Other tools and libraries may or may not be compatible - check their documentation if in doubt.
+
+## Explanation of the new k-quant methods
+
+The new methods available are:
+* 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)
+* 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.
+* 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.
+* GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
+* 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
+* GGML_TYPE_Q8_K - "type-0" 8-bit quantization. Only used for quantizing intermediate results. The difference to the existing Q8_0 is that the block size is 256. All 2-6 bit dot products are implemented for this quantization type.
 
+Refer to the Provided Files table below to see what files use which methods, and how.
 <!-- compatibility_ggml end -->
 
 ## Provided files
 | Name | Quant method | Bits | Size | Max RAM required | Use case |
 | ---- | ---- | ---- | ---- | ---- | ----- |
-| Falcon-40b-Instruct.ggmlv3.q4_0.bin | q4_0 | 4 | 23.54 GB | 26.04 GB | 4-bit. |
-| Falcon-40b-Instruct.ggmlv3.q4_1.bin | q4_1 | 4 | 26.15 GB | 28.65 GB | 4-bit. Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models. |
-| Falcon-40b-Instruct.ggmlv3.q5_0.bin | q5_0 | 5 | 28.77 GB | 31.27 GB | 5-bit. Higher accuracy, higher resource usage and slower inference. |
-| Falcon-40b-Instruct.ggmlv3.q5_1.bin | q5_1 | 5 | 31.38 GB | 33.88 GB | 5-bit. Even higher accuracy, resource usage and slower inference. |
+| falcon40b-instruct.ggmlv3.q2_K.bin | q2_K | 2 | 13.74 GB | 16.24 GB | New k-quant method. Uses GGML_TYPE_Q4_K for the attention.vw and feed_forward.w2 tensors, GGML_TYPE_Q2_K for the other tensors. |
+| falcon40b-instruct.ggmlv3.q3_K_L.bin | q3_K_L | 3 | 17.98 GB | 20.48 GB | New k-quant method. Uses GGML_TYPE_Q5_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else GGML_TYPE_Q3_K |
+| falcon40b-instruct.ggmlv3.q3_K_M.bin | q3_K_M | 3 | 17.98 GB | 20.48 GB | New k-quant method. Uses GGML_TYPE_Q4_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else GGML_TYPE_Q3_K |
+| falcon40b-instruct.ggmlv3.q3_K_S.bin | q3_K_S | 3 | 17.98 GB | 20.48 GB | New k-quant method. Uses GGML_TYPE_Q3_K for all tensors |
+| falcon40b-instruct.ggmlv3.q4_K_M.bin | q4_K_M | 4 | 23.54 GB | 26.04 GB | New k-quant method. Uses GGML_TYPE_Q6_K for half of the attention.wv and feed_forward.w2 tensors, else GGML_TYPE_Q4_K |
+| falcon40b-instruct.ggmlv3.q4_K_S.bin | q4_K_S | 4 | 23.54 GB | 26.04 GB | New k-quant method. Uses GGML_TYPE_Q4_K for all tensors |
+| falcon40b-instruct.ggmlv3.q5_K_M.bin | q5_K_M | 5 | 28.77 GB | 31.27 GB | New k-quant method. Uses GGML_TYPE_Q6_K for half of the attention.wv and feed_forward.w2 tensors, else GGML_TYPE_Q5_K |
+| falcon40b-instruct.ggmlv3.q5_K_S.bin | q5_K_S | 5 | 28.77 GB | 31.27 GB | New k-quant method. Uses GGML_TYPE_Q5_K for all tensors |
+| falcon40b-instruct.ggmlv3.q6_K.bin | q6_K | 6 | 34.33 GB | 36.83 GB | New k-quant method. Uses GGML_TYPE_Q8_K - 6-bit quantization - for all tensors |
+| falcon40b-instruct.ggmlv3.q8_0.bin | q8_0 | 8 | 44.46 GB | 46.96 GB | Original llama.cpp quant method, 8-bit. Almost indistinguishable from float16. High resource use and slow. Not recommended for most users. |
+
+
+**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.
+
+## How to run in `llama.cpp`
+
+I use the following command line; adjust for your tastes and needs:
+
+```
+./main -t 10 -ngl 32 -m gpt4-x-alpaca-13b.ggmlv3.q5_0.bin --color -c 2048 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "### Instruction: Write a story about llamas\n### Response:"
+```
+Change `-t 10` to the number of physical CPU cores you have. For example if your system has 8 cores/16 threads, use `-t 8`.
+
+Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
+
+If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
+
+## How to run in `text-generation-webui`
 
-A q8_0 file will be provided shortly. There is currently an issue preventing it from working. Once this is fixed, it will be uploaded.
+Further instructions here: [text-generation-webui/docs/llama.cpp-models.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/llama.cpp-models.md).
 
 <!-- footer start -->
 ## Discord
@@ -85,7 +118,7 @@ Donaters will get priority support on any and all AI/LLM/model questions and req
 
 **Special thanks to**: Luke from CarbonQuill, Aemon Algiz, Dmitriy Samsonov.
 
-**Patreon special mentions**: Oscar Rangel, Eugene Pentland, Talal Aujan, Cory Kujawski, Luke, Asp the Wyvern, Ai Maven, Pyrater, Alps Aficionado, senxiiz, Willem Michiel, Junyu Yang, trip7s trip, Sebastain Graf, Joseph William Delisle, Lone Striker, Jonathan Leane, Johann-Peter Hartmann, David Flickinger, Spiking Neurons AB, Kevin Schuppel, Mano Prime, Dmitriy Samsonov, Sean Connelly, Nathan LeClaire, Alain Rossmann, Fen Risland, Derek Yates, Luke Pendergrass, Nikolai Manek, Khalefa Al-Ahmad, Artur Olbinski, John Detwiler, Ajan Kanaga, Imad Khwaja, Trenton Dambrowitz, Kalila, vamX, webtim, Illia Dulskyi.
+**Patreon special mentions**: vamX, K, Jonathan Leane, Lone Striker, Sean Connelly, Chris McCloskey, WelcomeToTheClub, Nikolai Manek, John Detwiler, Kalila, David Flickinger, Fen Risland, subjectnull, Johann-Peter Hartmann, Talal Aujan, John Villwock, senxiiz, Khalefa Al-Ahmad, Kevin Schuppel, Alps Aficionado, Derek Yates, Mano Prime, Nathan LeClaire, biorpg, trip7s trip, Asp the Wyvern, chris gileta, Iucharbius , Artur Olbinski, Ai Maven, Joseph William Delisle, Luke Pendergrass, Illia Dulskyi, Eugene Pentland, Ajan Kanaga, Willem Michiel, Space Cruiser, Pyrater, Preetika Verma, Junyu Yang, Oscar Rangel, Spiking Neurons AB, Pierre Kircher, webtim, Cory Kujawski, terasurfer , Trenton Dambrowitz, Gabriel Puliatti, Imad Khwaja, Luke.
 
 Thank you to all my generous patrons and donaters!