English | 简体中文

👋 join us on Twitter, Discord and WeChat

---

News 🎉

• [2023/07] TurboMind supports Llama-2 70B with GQA.
• [2023/07] TurboMind supports Llama-2 7B/13B.
• [2023/07] TurboMind supports tensor-parallel inference of InternLM.

---

Introduction

LMDeploy is a toolkit for compressing, deploying, and serving LLM, developed by the MMRazor and MMDeploy teams. It has the following core features:

• Efficient Inference Engine (TurboMind): Based on FasterTransformer, we have implemented an efficient inference engine - TurboMind, which supports the inference of LLaMA and its variant models on NVIDIA GPUs.

• Interactive Inference Mode: By caching the k/v of attention during multi-round dialogue processes, it remembers dialogue history, thus avoiding repetitive processing of historical sessions.

• Multi-GPU Model Deployment and Quantization: We provide comprehensive model deployment and quantification support, and have been validated at different scales.

• Persistent Batch Inference: Further optimization of model execution efficiency.

Performance

Case I: output token throughput with fixed input token and output token number (1, 2048)

Case II: request throughput with real conversation data

Test Setting: LLaMA-7B, NVIDIA A100(80G)

The output token throughput of TurboMind exceeds 2000 tokens/s, which is about 5% - 15% higher than DeepSpeed overall and outperforms huggingface transformers by up to 2.3x. And the request throughput of TurboMind is 30% higher than vLLM.

Quick Start

Installation

Install lmdeploy with pip ( python 3.8+) or from source

pip install lmdeploy

Deploy InternLM

Get InternLM model

# 1. Download InternLM model

# Make sure you have git-lfs installed (https://git-lfs.com)
git lfs install
git clone https://huggingface.co/internlm/internlm-chat-7b /path/to/internlm-chat-7b

# if you want to clone without large files – just their pointers
# prepend your git clone with the following env var:
GIT_LFS_SKIP_SMUDGE=1

# 2. Convert InternLM model to turbomind's format, which will be in "./workspace" by default
python3 -m lmdeploy.serve.turbomind.deploy internlm-chat-7b /path/to/internlm-chat-7b

Inference by TurboMind

python -m lmdeploy.turbomind.chat ./workspace

Note
When inferring with FP16 precision, the InternLM-7B model requires at least 15.7G of GPU memory overhead on TurboMind.
It is recommended to use NVIDIA cards such as 3090, V100, A100, etc. Disable GPU ECC can free up 10% memory, try sudo nvidia-smi --ecc-config=0 and reboot system.

Note
Tensor parallel is available to perform inference on multiple GPUs. Add --tp=<num_gpu> on chat to enable runtime TP.

Serving with gradio

python3 -m lmdeploy.serve.gradio.app ./workspace

Serving with Triton Inference Server

Launch inference server by:

bash workspace/service_docker_up.sh

Then, you can communicate with the inference server by command line,

python3 -m lmdeploy.serve.client {server_ip_addresss}:33337

or webui,

python3 -m lmdeploy.serve.gradio.app {server_ip_addresss}:33337

For the deployment of other supported models, such as LLaMA, LLaMA-2, vicuna and so on, you can find the guide from here

Inference with PyTorch

For detailed instructions on Inference pytorch models, see here.

Single GPU

python3 -m lmdeploy.pytorch.chat $NAME_OR_PATH_TO_HF_MODEL \
    --max_new_tokens 64 \
    --temperture 0.8 \
    --top_p 0.95 \
    --seed 0

Tensor Parallel with DeepSpeed

deepspeed --module --num_gpus 2 lmdeploy.pytorch.chat \
    $NAME_OR_PATH_TO_HF_MODEL \
    --max_new_tokens 64 \
    --temperture 0.8 \
    --top_p 0.95 \
    --seed 0

You need to install deepspeed first to use this feature.

pip install deepspeed

Quantization

In fp16 mode, kv_cache int8 quantization can be enabled, and a single card can serve more users. First execute the quantization script, and the quantization parameters are stored in the workspace/triton_models/weights transformed by deploy.py.

python3 -m lmdeploy.lite.apis.kv_qparams \
  --model $HF_MODEL \
  --output_dir $DEPLOY_WEIGHT_DIR \
  --symmetry True \   # Whether to use symmetric or asymmetric quantization.
  --offload  False \  # Whether to offload some modules to CPU to save GPU memory.
  --num_tp 1 \   # The number of GPUs used for tensor parallelism

Then adjust workspace/triton_models/weights/config.ini

• use_context_fmha changed to 0, means off
• quant_policy is set to 4. This parameter defaults to 0, which means it is not enabled

Here is quantization test results.

Warning
runtime Tensor Parallel for quantilized model is not available. Please setup --tp on deploy to enable static TP.

Contributing

We appreciate all contributions to LMDeploy. Please refer to CONTRIBUTING.md for the contributing guideline.

Acknowledgement

• FasterTransformer

License

This project is released under the Apache 2.0 license.