Skip to content

Latest commit

 

History

History
278 lines (198 loc) · 7.45 KB

README.md

File metadata and controls

278 lines (198 loc) · 7.45 KB
Raw

TuGraph-DB ChatBot

TuGraph-DB ChatBot is a demo demonstrates the effect of an agent trained by the corpus generated by the Awesome-Text2GQL. It can intercract with the TuGraph-DB taking how you want to operate the db in Chinese as input, such as querying or creating data. And The ChatBot will also help you to excute the cypher too. Let's try it now!

demo

Requirements

This README.md has been tested in:

  • GPU: V100 16G-RAM

  • OS: CentOS7

  • Disk: System Disk 110G or larger (minimum requirements: System Disk 40G, Data Disk 60G, if install data and model in data disk)

Quick Start

1. Preparation

1.1 Install Nvidia Docker

Nvidia Docker allows you to use NVIDIA GPU in docker. Here is the installation method to install Nvidia Docker in CentOS7. You can install Nvidia Docker in ubuntu system by yourself as well. Make sure you have installed Nvidia Docker before you start the container of tugraph-db-runtime image in the following steps.

distribution=$(. /etc/os-release;echo$ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.repo | sudo tee /etc/yum.repos.d/nvidia-docker.repo

sudo yum install -y nvidia-docker2
sudo pkill -SIGHUP dockerd

Restart docker

sudo systemctl restart docker

1.2 Build Dependencies in Compile Container

Pull tugraph-db docker image and build it from source.

docker pull tugraph/tugraph-compile-centos7

Set environment variables

export VERSION=latest                                       # the version of compile image is latest
export REPOSITORY=docker.io/tugraph/tugraph-compile-centos7 # path to your docker image

Start Container

docker run -dt --name demo_comp ${REPOSITORY}:${VERSION} /bin/bash
Change the defalut python3.6 into python3.10

Python3.10 is required for the demo.

# install dependency package
yum groupinstall "Development Tools"
yum install openssl-devel bzip2-devel libffi-devel

# bulid Python3.10 from source
wget https://www.python.org/ftp/python/3.10.15/Python-3.10.15.tgz
tar xzf Python-3.10.15.tgz
cd Python-3.10.15
./configure --enable-optimizations --with-openssl=/usr/include/openssl
make altinstall

# verify
python3.10 --version

Set python3.10 as the default python.

update-alternatives --install /usr/local/bin/python3 python3 /usr/local/bin/python3.6 1
update-alternatives --install /usr/local/bin/python3 python3 /usr/local/bin/python3.10 2
update-alternatives --config python3

In the last step, you will see a list,choose python3.10 as the default python version.

Verify python version

python3 --version

Install cython

python3 -m pip install -i https://mirrors.aliyun.com/pypi/simple/ Cython

# or build from source (deprecated)

wget https://github.com/cython/cython/archive/refs/tags/3.0.11.tar.gz
tar xzf 3.0.11.tar.gz
cd cython-3.0.11
python3 setup.py install
Build TuGraph-DB From Source
mkdir work_repo && cd work_repo
git clone --recursive https://github.com/TuGraph-family/tugraph-db.git
cd tugraph-db
deps/build_deps.sh
mkdir build && cd build
cmake .. -DOURSYSTEM=centos7
make
make package
Copy Dependencies

Exit comple container.

exit

Copy dydamic link library to a certain path.

cd .
mkdir demo
docker cp demo_comp:/root/work_repo/tugraph-db/build/output/liblgraph_client_cpp_rpc.so  ./demo/
docker cp demo_comp:/root/work_repo/tugraph-db/build/output/liblgraph_client_python.so  ./demo/
docker cp demo_comp:/root/work_repo/tugraph-db/build/output/liblgraph.so  ./demo/

1.3 Prepare Runtime Container

Pull tugraph-db docker image and build it from source.

docker pull tugraph/tugraph-runtime-centos7

Set environment variables

export VERSION=latest
export REPOSITORY=docker.io/tugraph/tugraph-runtime-centos7 # path to your docker image

Start Container

docker run -dt --gpus all -p 7070:7070  -p 7687:7687 -p 9090:9090 -v /root/tugraph/data:/var/lib/lgraph/data  -v /root/tugraph/log:/var/log/lgraph_log \
-v $HOME/demo:/root/work_repo/demo \
 --name demo ${REPOSITORY}:${VERSION} /bin/bash

docker exec -it demo bash

#--gpus means enabling gpus in the docker container

Verify you can use gpu in the docker container. You will see the information of gpu after input the instruction bellow. If you can't see the gpu information, back to the the steps before to make sure you have installed Nvidia-Docker successfully and started docker with --gpus.

nvidia-smi

1.4 Generate Dataset In Runtime Container

Install MiniConda
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

bash Miniconda3-latest-Linux-x86_64.sh

# verify
conda --version

conda create -n demo python=3.10

Restart your shell or terminal to make your settings take effect.

conda activate demo
Generate Dataset with Awesome-Text2GQL
cd
mkdir -p /root/work_repo/
cd work_repo
git clone https://github.com/TuGraph-family/Awesome-Text2GQL.git
git clone https://github.com/eosphoros-ai/DB-GPT-Hub.git
git clone https://github.com/TuGraph-family/tugraph-db.git
cd /root/work_repo/Awesome-Text2GQL
mkdir tugraph-db

Add templates in ./input_examples/corpus_template.txt, and run the whole flow to get datasets generated after setting up the environments.Please check README.md for more details.

Finally, split the datasets into train and dev part. The datasets should be organized as bellow.

tugraph-db/
|-train.json                     # for training
|-dev.json                       # for evaluation
|-gold_dev.txt                   # for evaluation

2. Fine-tune

Before you excute training、inference or evaluation, mkdir for outputs:

cd ../DB-GPT-Hub/src/dbgpt-hub-gql/dbgpt_hub_gql
mkdir -p ./output/logs ./output/adapter ./output/pred

Copy datasets

cd /root/work_repo/DB-GPT-Hub/src/dbgpt-hub-gql/dbgpt-hub-gql
cp -r /path/to/datasets/generated/tugraph-db ./data  # path to your tugraph-db dataset built above.

Follow the README.md in DB-GPT-GQL to fine-tune LLMs. Replace the virtual env dbgpt_hub_gql with demo while installing dependencies. The model CodeLlama-7B-Instruct and the corresponding LoRA method are tested.

3. Run Demo

Copy necesarry dependencies into the demo directory.

cd /root/work_repo/
mkdir -p ./DB-GPT-Hub/src/dbgpt-hub-gql/dbgpt_hub_gql/demo/

cd /root/work_repo/demo
cp * /root/work_repo/DB-GPT-Hub/src/dbgpt-hub-gql/dbgpt_hub_gql/demo/

cd /root/work_repo/Awesome-Text2GQL/demo
cp * /root/work_repo/DB-GPT-Hub/src/dbgpt-hub-gql/dbgpt_hub_gql/demo/

cd /root/work_repo/
cp -r ./tugraph-db/demo/movie ./DB-GPT-Hub/src/dbgpt-hub-gql/dbgpt_hub_gql/demo/

cd DB-GPT-Hub/src/dbgpt-hub-gql/dbgpt_hub_gql/demo

Import data of TuGraph-DB

lgraph_import --dir /var/lib/lgraph/data --verbose 2 -c ./movie/import.json --continue_on_error 1 --overwrite 1 --online false
rm -rf import_tmp

Start TuGraph-DB

lgraph_server -c /usr/local/etc/lgraph.json -d run --log_dir ""

Run demo

cd ..
cd ..
sh ./dbgpt_hub_gql/demo/demo.sh

Stop server Ctrl+C