Edit model card
TheBlokeAI

OpenAccess AI Collective's Hippogriff 30B Chat GGML

These files are GGML format model files for OpenAccess AI Collective's Hippogriff 30B Chat.

GGML files are for CPU + GPU inference using llama.cpp and libraries and UIs which support this format, such as:

Repositories available

Prompt template

You are a helpful assistant
USER: prompt goes here
ASSISTANT:

or

<|system|> You are a helpful assistant
<|user|> prompt goes here
<|model|>

Compatibility

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.

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 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
hippogriff-30b.ggmlv3.q2_K.bin q2_K 2 13.60 GB 16.10 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.
hippogriff-30b.ggmlv3.q3_K_L.bin q3_K_L 3 17.20 GB 19.70 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
hippogriff-30b.ggmlv3.q3_K_M.bin q3_K_M 3 15.64 GB 18.14 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
hippogriff-30b.ggmlv3.q3_K_S.bin q3_K_S 3 13.98 GB 16.48 GB New k-quant method. Uses GGML_TYPE_Q3_K for all tensors
hippogriff-30b.ggmlv3.q4_0.bin q4_0 4 18.30 GB 20.80 GB Original llama.cpp quant method, 4-bit.
hippogriff-30b.ggmlv3.q4_1.bin q4_1 4 20.33 GB 22.83 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.
hippogriff-30b.ggmlv3.q4_K_M.bin q4_K_M 4 19.57 GB 22.07 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
hippogriff-30b.ggmlv3.q4_K_S.bin q4_K_S 4 18.30 GB 20.80 GB New k-quant method. Uses GGML_TYPE_Q4_K for all tensors
hippogriff-30b.ggmlv3.q5_0.bin q5_0 5 22.37 GB 24.87 GB Original llama.cpp quant method, 5-bit. Higher accuracy, higher resource usage and slower inference.
hippogriff-30b.ggmlv3.q5_1.bin q5_1 5 24.40 GB 26.90 GB Original llama.cpp quant method, 5-bit. Even higher accuracy, resource usage and slower inference.
hippogriff-30b.ggmlv3.q5_K_M.bin q5_K_M 5 23.02 GB 25.52 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
hippogriff-30b.ggmlv3.q5_K_S.bin q5_K_S 5 22.37 GB 24.87 GB New k-quant method. Uses GGML_TYPE_Q5_K for all tensors
hippogriff-30b.ggmlv3.q6_K.bin q6_K 6 26.69 GB 29.19 GB New k-quant method. Uses GGML_TYPE_Q8_K - 6-bit quantization - for all tensors
hippogriff-30b.ggmlv3.q8_0.bin q8_0 8 34.56 GB 37.06 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 hippogriff-30b.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

Further instructions here: text-generation-webui/docs/llama.cpp-models.md.

Discord

For further support, and discussions on these models and AI in general, join us at:

TheBloke AI's Discord server

Thanks, and how to contribute.

Thanks to the chirper.ai team!

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: OpenAccess AI Collective's Hippogriff 30B Chat

Hippogriff 30B Chat

Built with Axolotl

Hippogriff 30B Chat is an experiment that builds on Manticore with new datasets, while removing a few more instruction and chat datasets. It also includes a de-duped subset of the Pygmalion dataset. It also removes all Alpaca style prompts using ### in favor of chat only style prompts using USER:,ASSISTANT: as well as pygmalion/metharme prompting using <|system|>, <|user|> and <|model|> tokens.

Questions, comments, feedback, looking to donate, or want to help? Reach out on our Discord or email [email protected]

Training Datasets

Hippogriff 30B Chat is a Llama 30B model fine-tuned on the following datasets

Hippogriff differs from Manticore as it does not use the WizardLM, WizardVicuna, Alpaca, or ShareGPT datasets.

Initial thoughts

Hippogriff follows instructions pretty well. It still struggles with anything that has to do with math. Prose is much better than manticore. Handles in-context QA much better.

Shoutouts

Special thanks to Nanobit for helping with Axolotl, TheBloke for quantizing these models are more accessible to all, 0x000011b for the RP dataset, and the support from everyone in our AI Discord community.

Demo

A Spaces demo is not provided for this release due to 30B models currently not fitting in VRAM.

Build

Hippogriff was built with Axolotl on 8xA100 80GB

  • 1.5 epochs taking approximately 12 hours. No further epochs will be released.

Bias, Risks, and Limitations

Hippogriff has not been aligned to human preferences with techniques like RLHF or deployed with in-the-loop filtering of responses like ChatGPT, so the model can produce problematic outputs (especially when prompted to do so). Hippogriff was fine-tuned from the base model LlaMa 30B, please refer to its model card's Limitations Section for relevant information.

The core functionality of Llama is taking a string of text and predicting the next token. While language models are widely used for tasks other than this, there are a lot of unknowns with this work. When prompting Hippogriff it is important to remember that the statistically most likely next token is often not the token that produces the most "accurate" text. Never depend upon Hippogriff to produce factually accurate output.

Llama was trained on the Pile, a dataset known to contain profanity, lewd, and otherwise abrasive language. Depending upon use case Hippogriff may produce socially unacceptable text. See Sections 5 and 6 of the Pile paper for a more detailed analysis of the biases in the Pile.

As with all language models, it is hard to predict in advance how Hippogriff will respond to particular prompts and offensive content may occur without warning. We recommend having a human curate or filter the outputs before releasing them, both to censor undesirable content and to improve the quality of the results.

Examples

TBD

USER: write an original short story of at least 300 words in the style of Zora Neale Hurston. Incorporate the following literary devices: Euphemism and Foreshadowing.
ASSISTANT: Once upon a time, there was a black girl named Rose, whose skin shined like a pearl beneath the hot sun. She lived with her mama and daddy on their farm, where they grew crops to feed the family and neighbors. On this particular day, Rose felt a tinge of excitement as she watched her parents work diligently in their fields. 'It's almost harvest season!' she thought, imagining the delicious meals that awaited them. But, soon after, it became apparent that something sinister lurked amidst the rows of fresh produce. 

As her father hoed around a vine-covered cucumber plant, he unearthed a peculiar rock. It was the size of an acorn but had a strange shape, unlike anything he or his wife had ever seen. "What have we here?" He said, turning to his wife with curiosity and excitement. She examined the rock for herself and declared, "Why, it looks like a piece of coal!" But as both mother and daughter looked closer, it seemed to be far more than a mere piece of coal. A sense of unease crept up Rose's spine when she noticed that the rock resembled the dark face of a baby in the womb. As Rose tried to suppress a scream, her mother gasped, "Sweet Jesus! Let's bury this thing before anyone sees!" And so, they quickly hid the stone back underground, hoping that nobody would ever discover their secret.
USER: tell me 5 words that start with "ex" and then the sum of 33 and 77
ASSISTANT: Excite, exotic, exhibit, excitable, excessive
USER: sally (a girl) has 3 brothers. Each brother has 2 sisters. how many sisters does sally have?
ASSISTANT: Sally has 3 brothers, and each brother has 2 sisters. To find out how many sisters Sally has in total, we need to multiply the number of bro
thers by the number of sisters per brother, then add 1 to account for Sally herself. So, 3 * 2 + 1 = 6 sisters in total
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference API
Inference API (serverless) has been turned off for this model.

Space using TheBloke/hippogriff-30b-chat-GGML 1