TheBloke/GodziLLa2-70B-GGML

Chat & support: TheBloke's Discord server

Want to contribute? TheBloke's Patreon page

TheBloke's LLM work is generously supported by a grant from andreessen horowitz (a16z)

---

GodziLLa2 70B - GGML

• Model creator: MayaPH
• Original model: GodziLLa2 70B

Description

This repo contains GGML format model files for MayaPH's GodziLLa2 70B.

Important note regarding GGML files.

The GGML format has now been superseded by GGUF. As of August 21st 2023, llama.cpp no longer supports GGML models. Third party clients and libraries are expected to still support it for a time, but many may also drop support.

Please use the GGUF models instead.

About GGML

GPU acceleration is now available for Llama 2 70B GGML files, with both CUDA (NVidia) and Metal (macOS). The following clients/libraries are known to work with these files, including with GPU acceleration:

• llama.cpp, commit e76d630 and later.
• text-generation-webui, the most widely used web UI.
• KoboldCpp, version 1.37 and later. A powerful GGML web UI, especially good for story telling.
• LM Studio, a fully featured local GUI with GPU acceleration for both Windows and macOS. Use 0.1.11 or later for macOS GPU acceleration with 70B models.
• llama-cpp-python, version 0.1.77 and later. A Python library with LangChain support, and OpenAI-compatible API server.
• ctransformers, version 0.2.15 and later. A Python library with LangChain support, and OpenAI-compatible API server.

Repositories available

• GPTQ models for GPU inference, with multiple quantisation parameter options.
• 2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference
• 2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference (deprecated)
• MayaPH's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions

Prompt template: Alpaca

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
{prompt}

### Response:

Compatibility

Works with llama.cpp commit e76d630 until August 21st, 2023

Will not work with llama.cpp after commit dadbed99e65252d79f81101a392d0d6497b86caa.

For compatibility with latest llama.cpp, please use GGUF files instead.

Or one of the other tools and libraries listed above.

To use in llama.cpp, you must add -gqa 8 argument.

For other UIs and libraries, please check the docs.

Explanation of the new k-quant methods

Click to see details

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
godzilla2-70b.ggmlv3.q2_K.bin	q2_K	2	28.59 GB	31.09 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.
godzilla2-70b.ggmlv3.q3_K_S.bin	q3_K_S	3	29.75 GB	32.25 GB	New k-quant method. Uses GGML_TYPE_Q3_K for all tensors
godzilla2-70b.ggmlv3.q3_K_M.bin	q3_K_M	3	33.04 GB	35.54 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
godzilla2-70b.ggmlv3.q3_K_L.bin	q3_K_L	3	36.15 GB	38.65 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
godzilla2-70b.ggmlv3.q4_0.bin	q4_0	4	38.87 GB	41.37 GB	Original quant method, 4-bit.
godzilla2-70b.ggmlv3.q4_K_S.bin	q4_K_S	4	38.87 GB	41.37 GB	New k-quant method. Uses GGML_TYPE_Q4_K for all tensors
godzilla2-70b.ggmlv3.q4_K_M.bin	q4_K_M	4	41.38 GB	43.88 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
godzilla2-70b.ggmlv3.q4_1.bin	q4_1	4	43.17 GB	45.67 GB	Original quant method, 4-bit. Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models.
godzilla2-70b.ggmlv3.q5_0.bin	q5_0	5	47.46 GB	49.96 GB	Original quant method, 5-bit. Higher accuracy, higher resource usage and slower inference.
godzilla2-70b.ggmlv3.q5_K_S.bin	q5_K_S	5	47.46 GB	49.96 GB	New k-quant method. Uses GGML_TYPE_Q5_K for all tensors
godzilla2-70b.ggmlv3.q5_K_M.bin	q5_K_M	5	48.75 GB	51.25 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

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

Make sure you are using llama.cpp from commit dadbed99e65252d79f81101a392d0d6497b86caa or earlier.

For compatibility with latest llama.cpp, please use GGUF files instead.

I use the following command line; adjust for your tastes and needs:

./main -t 10 -ngl 40 -gqa 8 -m godzilla2-70b.ggmlv3.q4_K_M.bin --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n### Instruction:\n{prompt}\n\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. If you are fully offloading the model to GPU, use -t 1

Change -ngl 40 to the number of GPU layers you have VRAM for. Use -ngl 100 to offload all layers to VRAM - if you have a 48GB card, or 2 x 24GB, or similar. Otherwise you can partially offload as many as you have VRAM for, on one or more GPUs.

If you want to have a chat-style conversation, replace the -p <PROMPT> argument with -i -ins

Remember the -gqa 8 argument, required for Llama 70B models.

Change -c 4096 to the desired sequence length for this model. For models that use RoPE, add --rope-freq-base 10000 --rope-freq-scale 0.5 for doubled context, or --rope-freq-base 10000 --rope-freq-scale 0.25 for 4x context.

For other parameters and how to use them, please refer to the llama.cpp documentation

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.

• Patreon: https://patreon.com/TheBlokeAI
• Ko-Fi: https://ko-fi.com/TheBlokeAI

Special thanks to: Aemon Algiz.

Patreon special mentions: Russ Johnson, J, alfie_i, Alex, NimbleBox.ai, Chadd, Mandus, Nikolai Manek, Ken Nordquist, ya boyyy, Illia Dulskyi, Viktor Bowallius, vamX, Iucharbius, zynix, Magnesian, Clay Pascal, Pierre Kircher, Enrico Ros, Tony Hughes, Elle, Andrey, knownsqashed, Deep Realms, Jerry Meng, Lone Striker, Derek Yates, Pyrater, Mesiah Bishop, James Bentley, Femi Adebogun, Brandon Frisco, SuperWojo, Alps Aficionado, Michael Dempsey, Vitor Caleffi, Will Dee, Edmond Seymore, usrbinkat, LangChain4j, Kacper Wikieł, Luke Pendergrass, John Detwiler, theTransient, Nathan LeClaire, Tiffany J. Kim, biorpg, Eugene Pentland, Stanislav Ovsiannikov, Fred von Graf, terasurfer, Kalila, Dan Guido, Nitin Borwankar, 阿明, Ai Maven, John Villwock, Gabriel Puliatti, Stephen Murray, Asp the Wyvern, danny, Chris Smitley, ReadyPlayerEmma, S_X, Daniel P. Andersen, Olakabola, Jeffrey Morgan, Imad Khwaja, Caitlyn Gatomon, webtim, Alicia Loh, Trenton Dambrowitz, Swaroop Kallakuri, Erik Bjäreholt, Leonard Tan, Spiking Neurons AB, Luke @flexchar, Ajan Kanaga, Thomas Belote, Deo Leter, RoA, Willem Michiel, transmissions 11, subjectnull, Matthew Berman, Joseph William Delisle, David Ziegler, Michael Davis, Johann-Peter Hartmann, Talal Aujan, senxiiz, Artur Olbinski, Rainer Wilmers, Spencer Kim, Fen Risland, Cap'n Zoog, Rishabh Srivastava, Michael Levine, Geoffrey Montalvo, Sean Connelly, Alexandros Triantafyllidis, Pieter, Gabriel Tamborski, Sam, Subspace Studios, Junyu Yang, Pedro Madruga, Vadim, Cory Kujawski, K, Raven Klaugh, Randy H, Mano Prime, Sebastain Graf, Space Cruiser

Thank you to all my generous patrons and donaters!

And thank you again to a16z for their generous grant.

Original model card: MayaPH's GodziLLa2 70B

Released August 11, 2023

Model Description

GodziLLa 2 70B is an experimental combination of various proprietary LoRAs from Maya Philippines and Guanaco LLaMA 2 1K dataset, with LLaMA 2 70B. This model's primary purpose is to stress test the limitations of composite, instruction-following LLMs and observe its performance with respect to other LLMs available on the Open LLM Leaderboard. This model debuted in the leaderboard at rank #4 (August 17, 2023) and operates under the Llama 2 license.

Open LLM Leaderboard Metrics

Metric	Value
MMLU (5-shot)	69.88
ARC (25-shot)	71.42
HellaSwag (10-shot)	87.53
TruthfulQA (0-shot)	61.54
Average	72.59

According to the leaderboard description, here are the benchmarks used for the evaluation:

• MMLU (5-shot) - a test to measure a text model’s multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more.
• AI2 Reasoning Challenge -ARC- (25-shot) - a set of grade-school science questions.
• HellaSwag (10-shot) - a test of commonsense inference, which is easy for humans (~95%) but challenging for SOTA models.
• TruthfulQA (0-shot) - a test to measure a model’s propensity to reproduce falsehoods commonly found online.

A detailed breakdown of the evaluation can be found here. Huge thanks to @thomwolf.

Leaderboard Highlights (as of August 17, 2023)

• Godzilla 2 70B debuts at 4th place worldwide in the Open LLM Leaderboard.
• Godzilla 2 70B ranks #3 in the ARC challenge.
• Godzilla 2 70B ranks #5 in the TruthfulQA benchmark.
• *Godzilla 2 70B beats GPT-3.5 (ChatGPT) in terms of average performance and the HellaSwag benchmark (87.53 > 85.5).
• *Godzilla 2 70B outperforms GPT-3.5 (ChatGPT) and GPT-4 on the TruthfulQA benchmark (61.54 for G2-70B, 47 for GPT-3.5, 59 for GPT-4).
• *Godzilla 2 70B is on par with GPT-3.5 (ChatGPT) on the MMLU benchmark (<0.12%).

*Based on a leaderboard clone with GPT-3.5 and GPT-4 included.

Reproducing Evaluation Results

*Instruction template taken from Platypus 2 70B instruct.

Install LM Evaluation Harness:

# clone repository
git clone https://github.com/EleutherAI/lm-evaluation-harness.git
# change to repo directory
cd lm-evaluation-harness
# check out the correct commit
git checkout b281b0921b636bc36ad05c0b0b0763bd6dd43463
# install
pip install -e .

ARC:

python main.py --model hf-causal-experimental --model_args pretrained=MayaPH/GodziLLa2-70B --tasks arc_challenge --batch_size 1 --no_cache --write_out --output_path results/G270B/arc_challenge_25shot.json --device cuda --num_fewshot 25

HellaSwag:

python main.py --model hf-causal-experimental --model_args pretrained=MayaPH/GodziLLa2-70B --tasks hellaswag --batch_size 1 --no_cache --write_out --output_path results/G270B/hellaswag_10shot.json --device cuda --num_fewshot 10

MMLU:

python main.py --model hf-causal-experimental --model_args pretrained=MayaPH/GodziLLa2-70B --tasks hendrycksTest-* --batch_size 1 --no_cache --write_out --output_path results/G270B/mmlu_5shot.json --device cuda --num_fewshot 5

TruthfulQA:

python main.py --model hf-causal-experimental --model_args pretrained=MayaPH/GodziLLa2-70B --tasks truthfulqa_mc --batch_size 1 --no_cache --write_out --output_path results/G270B/truthfulqa_0shot.json --device cuda

Prompt Template

### Instruction:

<prompt> (without the <>)

### Response:

Technical Considerations

When using GodziLLa 2 70B, kindly take note of the following:

• The default precision is fp32, and the total file size that would be loaded onto the RAM/VRAM is around 275 GB. Consider using a lower precision (fp16, int8, int4) to save memory.
• To further save on memory, set the low_cpu_mem_usage argument to True.
• If you wish to use a quantized version of GodziLLa2-70B, you can either access TheBloke's GPTQ or GGML version of GodziLLa2-70B.
  • GodziLLa2-70B-GPTQ is available in 4-bit and 3-bit
  • GodziLLa2-70B-GGML is available in 8-bit, 6-bit, 5-bit, 4-bit, 3-bit, and 2-bit

Ethical Considerations

When using GodziLLa 2 70B, it is important to consider the following ethical considerations:

1. Privacy and Security: Avoid sharing sensitive personal information while interacting with the model. The model does not have privacy safeguards, so exercise caution when discussing personal or confidential matters.

2. Fairness and Bias: The model's responses may reflect biases present in the training data. Be aware of potential biases and make an effort to evaluate responses critically and fairly.

3. Transparency: The model operates as a predictive text generator based on patterns learned from the training data. The model's inner workings and the specific training data used are proprietary and not publicly available.

4. User Responsibility: Users should take responsibility for their own decisions and not solely rely on the information provided by the model. Consult with the appropriate professionals or reliable sources for specific advice or recommendations.

5. NSFW Content: The model is a merge of various datasets and LoRA adapters. It is highly likely that the resulting model contains uncensored content that may include, but is not limited to, violence, gore, explicit language, and sexual content. If you plan to further refine this model for safe/aligned usage, you are highly encouraged to implement guardrails along with it.

Further Information

For additional information or inquiries about GodziLLa 2 70B, please contact the Maya Philippines iOps Team via jasper.catapang@maya.ph.

Disclaimer

GodziLLa 2 70B is an AI language model from Maya Philippines. It is provided "as is" without warranty of any kind, express or implied. The model developers and Maya Philippines shall not be liable for any direct or indirect damages arising from the use of this model.

Acknowledgments

The development of GodziLLa 2 70B was made possible by Maya Philippines and the curation of the various proprietary datasets and creation of the different proprietary LoRA adapters. Special thanks to mlabonne for the Guanaco dataset found here. Last but not least, huge thanks to TheBloke for the quantized models, making our model easily accessible to a wider community.