Guanaco 7B ggml

From: https://huggingface.co/timdettmers/guanaco-7b

Original llama.cpp quant methods: `q4_0, q4_1, q5_0, q5_1, q8_0`

Quantized using an older version of llama.cpp and compatible with llama.cpp from May 19, commit 2d5db48.

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`

Quantization methods compatible with latest llama.cpp from June 6, commit 2d43387.

Files

Name	Quant method	Bits	Size	Max RAM required, no GPU offloading	Use case
guanaco-7B.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.
guanaco-7B.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
guanaco-7B.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
guanaco-7B.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
guanaco-7B.ggmlv3.q4_0.bin	q4_0	4	3.79 GB	6.29 GB	Original llama.cpp quant method, 4-bit.
guanaco-7B.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.
guanaco-7B.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
guanaco-7B.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
guanaco-7B.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.
guanaco-7B.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.
guanaco-7B.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
guanaco-7B.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
guanaco-7B.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
guanaco-7B.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.

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

Size	Dataset	Batch Size	Learning Rate	Max Steps	Sequence length
7B	OASST1	16	2e-4	1875	512
13B	OASST1	16	2e-4	1875	512
33B	OASST1	16	1e-4	1875	512
65B	OASST1	16	1e-4	1875	512

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.

Benchmark	Vicuna		Vicuna		OpenAssistant		-
Prompts	80		80		953
Judge	Human		GPT-4		GPT-4
Model	Elo	Rank	Elo	Rank	Elo	Rank	Median Rank
GPT-4	1176	1	1348	1	1294	1	1
Guanaco-65B	1023	2	1022	2	1008	3	2
Guanaco-33B	1009	4	992	3	1002	4	4
ChatGPT-3.5 Turbo	916	7	966	5	1015	2	5
Vicuna-13B	984	5	974	4	936	5	5
Guanaco-13B	975	6	913	6	885	6	6
Guanaco-7B	1010	3	879	8	860	7	7
Bard	909	8	902	7	-	-	8

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.

Dataset	7B	13B	33B	65B
LLaMA no tuning	35.1	46.9	57.8	63.4
Self-Instruct	36.4	33.3	53.0	56.7
Longform	32.1	43.2	56.6	59.7
Chip2	34.5	41.6	53.6	59.8
HH-RLHF	34.9	44.6	55.8	60.1
Unnatural Instruct	41.9	48.1	57.3	61.3
OASST1 (Guanaco)	36.6	46.4	57.0	62.2
Alpaca	38.8	47.8	57.3	62.5
FLAN v2	44.5	51.4	59.2	63.9

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.

	LLaMA-65B	GPT-3	OPT-175B	Guanaco-65B
Gender	70.6	62.6	65.7	47.5
Religion	{79.0}	73.3	68.6	38.7
Race/Color	57.0	64.7	68.6	45.3
Sexual orientation	{81.0}	76.2	78.6	59.1
Age	70.1	64.4	67.8	36.3
Nationality	64.2	61.6	62.9	32.4
Disability	66.7	76.7	76.7	33.9
Physical appearance	77.8	74.6	76.2	43.1
Socioeconomic status	71.5	73.8	76.2	55.3
Average	66.6	67.2	69.5	43.5

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}
}