Chat & support: my new Discord server

Want to contribute? TheBloke's Patreon page

Tim Dettmers' Guanaco 65B GPTQ

These files are GPTQ model files for Tim Dettmers' Guanaco 65B .

Multiple GPTQ parameter permutations are provided; see Provided Files below for details of the options provided, their parameters, and the software used to create them.

These models were quantised using hardware kindly provided by Latitude.sh .

Repositories available

• GPTQ models for GPU inference, with multiple quantisation parameter options.
• 2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference
• Unquantised fp16 model in pytorch format, for GPU inference and for further conversions

Prompt template: Guanaco

### Human: {prompt}
### Assistant:

Provided files

Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements.

Each separate quant is in a different branch. See below for instructions on fetching from different branches.

How to download from branches

• In text-generation-webui, you can add :branch to the end of the download name, eg TheBloke/guanaco-65B-GPTQ:gptq-4bit-32g-actorder_True
• With Git, you can clone a branch with:

git clone --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/guanaco-65B-GPTQ`

• In Python Transformers code, the branch is the revision parameter; see below.

How to easily download and use this model in text-generation-webui .

Please make sure you're using the latest version of text-generation-webui .

It is strongly recommended to use the text-generation-webui one-click-installers unless you know how to make a manual install.

• To download from a specific branch, enter for example TheBloke/guanaco-65B-GPTQ:gptq-4bit-32g-actorder_True
• see Provided Files above for the list of branches for each option.

• Note that you do not need to set GPTQ parameters any more. These are set automatically from the file quantize_config.json .

How to use this GPTQ model from Python code

First make sure you have AutoGPTQ installed:

GITHUB_ACTIONS=true pip install auto-gptq

Then try the following example code:

from transformers import AutoTokenizer, pipeline, logging
from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig

model_name_or_path = "TheBloke/guanaco-65B-GPTQ"
model_basename = "Guanaco-65B-GPTQ-4bit-128g.no-act.order"

use_triton = False

tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)

model = AutoGPTQForCausalLM.from_quantized(model_name_or_path,
        model_basename=model_basename
        use_safetensors=True,
        trust_remote_code=True,
        device="cuda:0",
        use_triton=use_triton,
        quantize_config=None)

"""
To download from a specific branch, use the revision parameter, as in this example:

model = AutoGPTQForCausalLM.from_quantized(model_name_or_path,
        revision="gptq-4bit-32g-actorder_True",
        model_basename=model_basename,
        use_safetensors=True,
        trust_remote_code=True,
        device="cuda:0",
        quantize_config=None)
"""

prompt = "Tell me about AI"
prompt_template=f'''### Human: {prompt}
### Assistant:
'''

print("\n\n*** Generate:")

input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
output = model.generate(inputs=input_ids, temperature=0.7, max_new_tokens=512)
print(tokenizer.decode(output[0]))

# Inference can also be done using transformers' pipeline

# Prevent printing spurious transformers error when using pipeline with AutoGPTQ
logging.set_verbosity(logging.CRITICAL)

print("*** Pipeline:")
pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_new_tokens=512,
    temperature=0.7,
    top_p=0.95,
    repetition_penalty=1.15
)

print(pipe(prompt_template)[0]['generated_text'])

Compatibility

The files provided will work with AutoGPTQ (CUDA and Triton modes), GPTQ-for-LLaMa (only CUDA has been tested), and Occ4m's GPTQ-for-LLaMa fork.

ExLlama works with Llama models in 4-bit. Please see the Provided Files table above for per-file compatibility.

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 : Luke from CarbonQuill, Aemon Algiz.

Patreon special mentions : Space Cruiser, Nikolai Manek, Sam, Chris McCloskey, Rishabh Srivastava, Kalila, Spiking Neurons AB, Khalefa Al-Ahmad, WelcomeToTheClub, Chadd, Lone Striker, Viktor Bowallius, Edmond Seymore, Ai Maven, Chris Smitley, Dave, Alexandros Triantafyllidis, Luke @flexchar, Elle, ya boyyy, Talal Aujan, Alex , Jonathan Leane, Deep Realms, Randy H, subjectnull, Preetika Verma, Joseph William Delisle, Michael Levine, chris gileta, K, Oscar Rangel, LangChain4j, Trenton Dambrowitz, Eugene Pentland, Johann-Peter Hartmann, Femi Adebogun, Illia Dulskyi, senxiiz, Daniel P. Andersen, Sean Connelly, Artur Olbinski, RoA, Mano Prime, Derek Yates, Raven Klaugh, David Flickinger, Willem Michiel, Pieter, Willian Hasse, vamX, Luke Pendergrass, webtim, Ghost , Rainer Wilmers, Nathan LeClaire, Will Dee, Cory Kujawski, John Detwiler, Fred von Graf, biorpg, Iucharbius , Imad Khwaja, Pierre Kircher, terasurfer , Asp the Wyvern, John Villwock, theTransient, zynix , Gabriel Tamborski, Fen Risland, Gabriel Puliatti, Matthew Berman, Pyrater, SuperWojo, Stephen Murray, Karl Bernard, Ajan Kanaga, Greatston Gnanesh, Junyu Yang.

Thank you to all my generous patrons and donaters!

Original model card: Tim Dettmers' Guanaco 65B

Guanaco Models Based on LLaMA

| Paper | Code | Demo |

The Guanaco models are open-source finetuned chatbots obtained through 4-bit QLoRA tuning of LLaMA base models on the OASST1 dataset. They are available in 7B, 13B, 33B, and 65B parameter sizes.

⚠️Guanaco is a model purely intended for research purposes and could produce problematic outputs.

Why use Guanaco?

• Competitive with commercial chatbot systems on the Vicuna and OpenAssistant benchmarks (ChatGPT and BARD) according to human and GPT-4 raters. We note that the relative performance on tasks not covered in these benchmarks could be very different. In addition, commercial systems evolve over time (we used outputs from the March 2023 version of the models).
• Available open-source for research purposes . Guanaco models allow cheap and local experimentation with high-quality chatbot systems.
• Replicable and efficient training procedure that can be extended to new use cases. Guanaco training scripts are available in the QLoRA repo .
• Rigorous comparison to 16-bit methods (both 16-bit full-finetuning and LoRA) in our paper demonstrates the effectiveness of 4-bit QLoRA finetuning.
• Lightweight checkpoints which only contain adapter weights.

License and Intended Use

Guanaco adapter weights are available under Apache 2 license. Note the use of the Guanaco adapter weights, requires access to the LLaMA model weighs. Guanaco is based on LLaMA and therefore should be used according to the LLaMA license.

Usage

Here is an example of how you would load Guanaco 7B in 4-bits:

import torch
from peft import PeftModel    
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig

model_name = "huggyllama/llama-7b"
adapters_name = 'timdettmers/guanaco-7b'

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    load_in_4bit=True,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    max_memory= {i: '24000MB' for i in range(torch.cuda.device_count())},
    quantization_config=BitsAndBytesConfig(
        load_in_4bit=True,
        bnb_4bit_compute_dtype=torch.bfloat16,
        bnb_4bit_use_double_quant=True,
        bnb_4bit_quant_type='nf4'
    ),
)
model = PeftModel.from_pretrained(model, adapters_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

Inference can then be performed as usual with HF models as follows:

prompt = "Introduce yourself"
formatted_prompt = (
    f"A chat between a curious human and an artificial intelligence assistant."
    f"The assistant gives helpful, detailed, and polite answers to the user's questions.\n"
    f"### Human: {prompt} ### Assistant:"
)
inputs = tokenizer(formatted_prompt, return_tensors="pt").to("cuda:0")
outputs = model.generate(inputs=inputs.input_ids, max_new_tokens=20)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Expected output similar to the following:

A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.
### Human: Introduce yourself ### Assistant: I am an artificial intelligence assistant. I am here to help you with any questions you may have.

Current Inference Limitations

Currently, 4-bit inference is slow. We recommend loading in 16 bits if inference speed is a concern. We are actively working on releasing efficient 4-bit inference kernels.

Below is how you would load the model in 16 bits:

model_name = "huggyllama/llama-7b"
adapters_name = 'timdettmers/guanaco-7b'
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    max_memory= {i: '24000MB' for i in range(torch.cuda.device_count())},
)
model = PeftModel.from_pretrained(model, adapters_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

Model Card

Architecture : The Guanaco models are LoRA adapters to be used on top of LLaMA models. They are added to all layers. For all model sizes, we use $r=64$.

Base Model : Guanaco uses LLaMA as base model with sizes 7B, 13B, 33B, 65B. LLaMA is a causal language model pretrained on a large corpus of text. See LLaMA paper for more details. Note that Guanaco can inherit biases and limitations of the base model.

Finetuning Data : Guanaco is finetuned on OASST1. The exact dataset is available at timdettmers/openassistant-guanaco .

Languages : The OASST1 dataset is multilingual (see the paper for details) and as such Guanaco responds to user queries in different languages. We note, however, that OASST1 is heavy in high-resource languages. In addition, human evaluation of Guanaco was only performed in English and based on qualitative analysis we observed degradation in performance in other languages.

Next, we describe Training and Evaluation details.

Training

Guanaco models are the result of 4-bit QLoRA supervised finetuning on the OASST1 dataset.

All models use NormalFloat4 datatype for the base model and LoRA adapters on all linear layers with BFloat16 as computation datatype. We set LoRA $r=64$, $\alpha=16$. We also use Adam beta2 of 0.999, max grad norm of 0.3 and LoRA dropout of 0.1 for models up to 13B and 0.05 for 33B and 65B models. For the finetuning process, we use constant learning rate schedule and paged AdamW optimizer.

Training hyperparameters

Evaluation

We test generative language capabilities through both automated and human evaluations. This second set of evaluations relies on queries curated by humans and aims at measuring the quality of model responses. We use the Vicuna and OpenAssistant datasets with 80 and 953 prompts respectively.

In both human and automated evaluations, for each prompt, raters compare all pairs of responses across the models considered. For human raters we randomize the order of the systems, for GPT-4 we evaluate with both orders.

We also use the MMLU benchmark to measure performance on a range of language understanding tasks. This is a multiple-choice benchmark covering 57 tasks including elementary mathematics, US history, computer science, law, and more. We report 5-shot test accuracy.

Risks and Biases

The model can produce factually incorrect output, and should not be relied on to produce factually accurate information. The model was trained on various public datasets; it is possible that this model could generate lewd, biased, or otherwise offensive outputs.

However, we note that finetuning on OASST1 seems to reduce biases as measured on the CrowS dataset. We report here the performance of Guanaco-65B compared to other baseline models on the CrowS dataset.

Citation

@article{dettmers2023qlora,
  title={QLoRA: Efficient Finetuning of Quantized LLMs},
  author={Dettmers, Tim and Pagnoni, Artidoro and Holtzman, Ari and Zettlemoyer, Luke},
  journal={arXiv preprint arXiv:2305.14314},
  year={2023}
}