Original llama.cpp quant methods: q4_0, q4_1, q5_0, q5_1, q8_0
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.
They should be compatible with all current UIs and libraries that use llama.cpp, such as those listed at the top of this README.
These new quantisation methods are only compatible with llama.cpp as of June 6th, commit 2d43387.
They will NOT be compatible with koboldcpp, text-generation-ui, and other UIs and libraries yet. Support is expected to come over the next few days.
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.
Provided files
Name
Quant method
Bits
Size
Max RAM required
Use case
vicuna-7b-1.1.ggmlv3.q2_K.bin
q2_K
2
2.80 GB
5.30 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.
vicuna-7b-1.1.ggmlv3.q3_K_L.bin
q3_K_L
3
3.55 GB
6.05 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
vicuna-7b-1.1.ggmlv3.q3_K_M.bin
q3_K_M
3
3.23 GB
5.73 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
vicuna-7b-1.1.ggmlv3.q3_K_S.bin
q3_K_S
3
2.90 GB
5.40 GB
New k-quant method. Uses GGML_TYPE_Q3_K for all tensors
vicuna-7b-1.1.ggmlv3.q4_0.bin
q4_0
4
3.79 GB
6.29 GB
Original llama.cpp quant method, 4-bit.
vicuna-7b-1.1.ggmlv3.q4_1.bin
q4_1
4
4.21 GB
6.71 GB
Original llama.cpp quant method, 4-bit. Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models.
vicuna-7b-1.1.ggmlv3.q4_K_M.bin
q4_K_M
4
4.05 GB
6.55 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
vicuna-7b-1.1.ggmlv3.q4_K_S.bin
q4_K_S
4
3.79 GB
6.29 GB
New k-quant method. Uses GGML_TYPE_Q4_K for all tensors
vicuna-7b-1.1.ggmlv3.q5_0.bin
q5_0
5
4.63 GB
7.13 GB
Original llama.cpp quant method, 5-bit. Higher accuracy, higher resource usage and slower inference.
vicuna-7b-1.1.ggmlv3.q5_1.bin
q5_1
5
5.06 GB
7.56 GB
Original llama.cpp quant method, 5-bit. Even higher accuracy, resource usage and slower inference.
vicuna-7b-1.1.ggmlv3.q5_K_M.bin
q5_K_M
5
4.77 GB
7.27 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
vicuna-7b-1.1.ggmlv3.q5_K_S.bin
q5_K_S
5
4.63 GB
7.13 GB
New k-quant method. Uses GGML_TYPE_Q5_K for all tensors
vicuna-7b-1.1.ggmlv3.q6_K.bin
q6_K
6
5.53 GB
8.03 GB
New k-quant method. Uses GGML_TYPE_Q8_K - 6-bit quantization - for all tensors
vicuna-7b-1.1.ggmlv3.q8_0.bin
q8_0
8
7.16 GB
9.66 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 vicuna-7b-1.1.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
I'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.
If 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.
Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
Special thanks to: Luke from CarbonQuill, Aemon Algiz, Dmitriy Samsonov.
Patreon special mentions: Ajan Kanaga, Kalila, Derek Yates, Sean Connelly, Luke, Nathan LeClaire, Trenton Dambrowitz, Mano Prime, David Flickinger, vamX, Nikolai Manek, senxiiz, Khalefa Al-Ahmad, Illia Dulskyi, trip7s trip, Jonathan Leane, Talal Aujan, Artur Olbinski, Cory Kujawski, Joseph William Delisle, Pyrater, Oscar Rangel, Lone Striker, Luke Pendergrass, Eugene Pentland, Johann-Peter Hartmann.
Thank you to all my generous patrons and donaters!
Original model card: LmSys' Vicuna 7B 1.1
NOTE: This "delta model" cannot be used directly.
Users have to apply it on top of the original LLaMA weights to get actual Vicuna weights.
See https://github.com/lm-sys/FastChat#vicuna-weights for instructions.
Vicuna Model Card
Model details
Model type:
Vicuna is an open-source chatbot trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.
It is an auto-regressive language model, based on the transformer architecture.
Model date:
Vicuna was trained between March 2023 and April 2023.
Organizations developing the model:
The Vicuna team with members from UC Berkeley, CMU, Stanford, and UC San Diego.
Primary intended uses:
The primary use of Vicuna is research on large language models and chatbots.
Primary intended users:
The primary intended users of the model are researchers and hobbyists in natural language processing, machine learning, and artificial intelligence.
Training dataset
70K conversations collected from ShareGPT.com.
Evaluation dataset
A preliminary evaluation of the model quality is conducted by creating a set of 80 diverse questions and utilizing GPT-4 to judge the model outputs. See https://vicuna.lmsys.org/ for more details.
Major updates of weights v1.1
Refactor the tokenization and separator. In Vicuna v1.1, the separator has been changed from "###" to the EOS token "</s>". This change makes it easier to determine the generation stop criteria and enables better compatibility with other libraries.
Fix the supervised fine-tuning loss computation for better model quality.
