LLM_Project

Main structure

LLM_project
│   docs
│   │   # literature review, project proposal, project report, etc.
|   codes
|   |   extract
|   |   |   # extracting info from music
|   |   process
|   |   |   # processing info
|   |   generate
|   |   |   # generating image
|   |   style_transfer
|   |   |   # transfering style
|   |   data
|   |   |   # data used in the project and output
|   |   main.ipynb
|   |   |   # ipynb for testing the pipeline
|   demo
|   |   docker
|   |   |   # docker files
|   |   generateAPI.py
|   |   generateUI.py
|   |   demo.py
|   |   |   # demo code
│   README.md
|   requirements.txt
|   environment.sh

Getting Started

Environment

bash environment.sh

Demo

To run the demo

python demo/generateAPI.py

python demo/generateUI.py

in two terminals, then open the browser and visit http://localhost:7860/ to see the demo

Docker

If you don't want to build a docker, you can change IngressHost and NameSpace in project_lyy.yaml into yours and use the docker we already built by running the following command.

cd demo/docker/MI-T_helm_template
make our
make connect

Before building the docker, please download pretrained model from https://download.pytorch.org/models/vgg19-dcbb9e9d.pth and put it in demo/docker/MI-T_API_docker/.

Build the docker:

docker login harbor.ai.iiis.co:9443
cd demo/docker/MI-T_API_docker
make docker USER=your_docker_username
make push USER=your_docker_username

cd demo/docker/MI-T_UI_docker
make docker USER=your_docker_username
make push USER=your_docker_username

You can remove local images after pushing.

make remove USER=your_docker_username

We already build the image for you. In codes/demo/docker/MI-T_helm_template/project.yaml, change your namespace to your own namespace, ContainerImage to your api image, GradioImage to your ui image, IngressHost to your own host.

To run the docker

cd demo/docker/MI-T_helm_template
make podup
make connect

Then you can open the browser and visit https://llmlab.ddns.net:9443 (Your ddns) to see the demo.

make poddown

remove the pod after you finish.

Detailed tasks

Extract

Two parts, from lyrics and from music. Qwen does both jobs well.

Input : music

Output : prompt

Process

Throwing everything to GPT/GLM.

Input : prompt

Output : prompt

Generate

Generate image using the prompt.

Input : prompt/dictionary

Output : image

Style Transfer

Transfer the style of the image. Make it more like an illustration or a cover.