gaopengpjlab / OmniQuant

An efficient, accurate, and omnibearing quantization algorithm for LLMs, encompassing both weight-only quantization (W4A16/W3A16/W2A16) and weight-activation quantization (W6A6, W4A4)

OmniQuant

An efficient, accurate, and omnibearing quantization algorithm for LLMs, encompassing both weight-only quantization (W4A16/W3A16/W2A16) and weight-activation quantization (W6A6, W4A4):

OmniQuant introduces optimization into quantization, but also keeps the data and time efficiency like PTQ. For example, OmniQuant can quantize LLaMa-2 model family (7B-70B) on a single A100-40G GPU within 1-16 hours using 128 samples.

The current release supports:

• OmniQuant algorithm for accurate weight-only quantization (W4A16/W3A16/W2A16) and weight-activation quantization (W6A6, W4A4)
• Pre-trained Omniquant model zoo for LLMs (LLaMA-1&2, LLaMA-2-Chat, OPT; load to generate quantized weights).
• A out-of-the-box case that leverages MLC-LLM to run LLaMa-2-Chat (7B/13B) with W3A16g128 quantization on GPUs and mobile phones.

Contents

• Install
• Omniquant Model Zoo
• Usage
• Inference with MLC-LLM
• Results
• Citation

Install

conda create -n omniquant python=3.10 -y
conda activate omniquant
git clone https://github.com/OpenGVLab/OmniQuant.git
cd OmniQuant
pip install --upgrade pip 
pip install -e .

OmniQuant Model Zoo

We provide pre-trained Omniquant model zoo for multiple model families, including LLaMa-1&2, LLaMa-2-Chat, OPT.

You can download the pre-trained model you nedd at Huggingface.

The detailed support list:

Models	Sizes	W2A16	W2A16g128	W2A16g64	W3A16
LLaMA	7B/13B/30B/65B	✅	✅	✅	✅
LLaMA-2	7B/13B/70B	✅	✅	✅	✅
OPT	125m/1.3B/2.7B/6.7B/13B/30B/66B	✅	✅	✅	✅

Models	Sizes	W3A16g128	W4A16	W4A16g128	W6A6	W4A4
LLaMA	7B/13B/30B/65B	✅	✅	✅	✅	✅
LLaMA-2	7B/13B/70B	✅	✅	✅	✅	✅
OPT	125m/1.3B/2.7B/6.7B/13B/30B/66B	✅	✅	✅	✅	✅
LLaMA-2-Chat	7B/13B	✅

Usage

We provide full script to run OmniQuant in ./scripts/. We use LLaMa-7B as an example here:

1. Obtain the channel-wise scales and shifts required for initialization:

conda install git git-lfs
git lfs install
git clone https://huggingface.co/ChenMnZ/act_shifts
git clone https://huggingface.co/ChenMnZ/act_scales

Optional, we also offer the script that you can generate channel-wise scales and shifts by yourself:

python generate_act_scale_shift.py --model /PATH/TO/LLaMA/llama-7b

2. Weight-only quantization

# W3A16
CUDA_VISIBLE_DEVICES=0 python main.py \
--model /PATH/TO/LLaMA/llama-7b  \
--epochs 20 --output_dir ./log/llama-7b-w3a16 \
--eval_ppl --wbits 3 --abits 16 --lwc

# W3A16g128
CUDA_VISIBLE_DEVICES=0 python main.py \
--model /PATH/TO/LLaMA/llama-7b  \
--epochs 20 --output_dir ./log/llama-7b-w3a16g128 \
--eval_ppl --wbits 3 --abits 16 --group_size 128 --lwc

3. weight-activation quantization

# W4A4
CUDA_VISIBLE_DEVICES=0 python main.py \
--model /PATH/TO/LLaMA/llama-7b  \
--epochs 20 --output_dir ./log/llama-7b-w4a4 \
--eval_ppl --wbits 4 --abits 4 --lwc --let \
--tasks piqa,arc_easy,arc_challenge,boolq,hellaswag,winogrande

4. save fake quantization models for further experiments. For example, mlc-llm compilation and GPT-4 evaluation.

CUDA_VISIBLE_DEVICES=0 python main.py \
--model /PATH/TO/LLaMA/llama-7b  \
--epochs 20 --output_dir ./log/llama-7b-w3a16g128 \
--wbits 3 --abits 16 --group_size 128 --lwc \
--save_dir /PATH/TO/SAVE/llama-7b-omniquant-w3a16g128

5. evaluate pre-trained OmniQuant models

CUDA_VISIBLE_DEVICES=0 python main.py \
--model /PATH/TO/LLaMA/llama-7b  \
--epochs 0 --output_dir ./log/llama-7b-w3a16g128 \
--eval_ppl --wbits 3 --abits 16 --group_size 128 --lwc \
--resume /PATH/TO/OmniQuant_Checkpoints/llama-7b-w3a16g128.pth

OmniQuant can quantized LLaMa-1&2(7B-70B) in one A100-40G GPU. However, we utilize fake quantization here, thereby can not reduce memory requirement when inference. To inference larger network please use --multigpu.

Runing Quantized Models with MLC-LLM

MLC-LLM offers a universal deployment solution suitable for various language models across a wide range of hardware backends, encompassing iPhones, Android phones, and GPUs from NVIDIA, AMD, and Intel.

We compile the OmniQuant's quantization models through MLC-LLM and offer an out-of-the-box case here. You can see smaller gpu memory usage and inference speedup:

conda install git git-lfs
git lfs install
mkdir dist && cd dist

# test Llama-2-7b-chat with w3a16g128 quantization
git clone https://huggingface.co/ChenMnZ/Llama-2-7b-chat-omniquant-w3a16g128asym
./mlc_chat_cli --local-id Llama-2-7b-chat-omniquant-w3a16g128asym --device-name cuda

# test Llama-2-13b-chat with w3a16g128 quantization
git clone https://huggingface.co/ChenMnZ/Llama-2-13b-chat-omniquant-w3a16g128asym
./mlc_chat_cli --local-id Llama-2-13b-chat-omniquant-w3a16g128asym --device-name cuda

Specially, we also deploy the aforementioned two quantized models into mobile phones through MLC-LLM. You can download the Android app by simply clicking the button below:

This app includes three models, LLaMa-2-7B-Chat-Omniquant-W3A16g128asym, LLaMa-2-13B-Chat-Omniquant-W3A16g128asym, and LLaMa-2-13B-Chat-Omniquant-W2A16g128asym. They require at least 4.5G, 7.5G, and 6.0G free RAM, respectively. Note that 2bit quantization has worse performance compared to 3bit quantization as shown in our paper. The inclusion of 2-bit quantization is just an extreme exploration about deploy LLM in mobile phones. Currently, this app is in its demo phase and may experience slower response times, so wait patiently for the generation of response. We have tested this app on Redmi Note 12 Turbo (Snapdragon 7+ Gen 2 and 16G RAM), some examples are provided below:

• LLaMa-2-7B-Chat-Omniquant-W3A16g128asym

• LLaMa-2-13B-Chat-Omniquant-W3A16g128asym

• LLaMa-2-13B-Chat-Omniquant-W2A16g128asym

Results

• OmniQuant achieve SoTA performance in weight-only quantization
• OmniQuant achieve SoTA performance in weight-activation quantization
• OmniQuant is generalize, also obatins excellent performance in instruction-tuned models with GPT-4 evaluation
• MLC-LLM can obtain really speedup and memory saving for W4A16/W3A16/W2A16 quantization

Related Project

SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models

AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration

GPTQ: Accurate Post-training Compression for Generative Pretrained Transformers

RPTQ: Reorder-Based Post-Training Quantization for Large Language Models

MLC LLM

Citation

If you use our OmniQuant approach in your research, please cite our paper:

@article{OmniQuant,
  title={OmniQuant: Omnidirectionally Calibrated Quantization for Large Language Models},
  author={Shao, Wenqi and Chen,Mengzhao and  Zhang, Zhaoyang and Xu, Peng and Zhao, Lirui and Li, Zhiqian and Zhang, Kaipeng Zhang, and Gao, Peng, and Qiao, Yu, and Luo, Ping},
  journal={arXiv preprint arXiv:2308.13137},
  year={2023}
}