VQAweb

Machine Type: n1-standard-8 vCPUs: 8 RAM: 30 GiB
GPU: NVIDIA T4 (1 unit)
Operating System: Ubuntu 22.04 LTS

A multimodal answering system based on RAG.

System Specifications

For a hassle-free installation, run the GCP-install-dependencies.sh script to install all required dependencies on a Ubuntu GCP instance.

This project requires .h5 model files to run properly. Follow these steps to set up the models:

Ensure git and git-lfs are installed

If you don’t have Git LFS installed, you can install it by running:
```
git lfs install
```
Navigate to the backend directory

Change into the VQAweb/backend directory where the models will be set up:
```
cd VQAweb/backend
```
Clone the VQAmodels repository

Clone the VQAmodels repository from Hugging Face:
```
git clone https://huggingface.co/930727fre/VQAmodels models
```
Move the .h5 files to the parent directory

Move the .h5 model files from the models directory to the VQAweb/backend directory:
```
mv models/*.h5 .
```
Remove the empty models directory

Once the files are moved, remove the now-empty models directory:
```
sudo rm -drf models
```

The stable version is located in the main branch.
Navigate to the project directory:
```
cd VQAweb
```
Modify the file VQAweb/frontend/Present/src/components/Pictureinput.vue:

Replace localhost in the axios.post line with your <server_IP>.
Run the backend using the following script:
```
./docker_run.sh
```
If you encounter execution issues, make the script executable first:
```
chmod +x ./docker_run.sh
./docker_run.sh
```
Access the application in your browser:

Visit <server_IP>:8000 in your web browser.
To stop the application:

Press Ctrl + C in the terminal.

Note:

If Docker images are not successfully deleted, manually modify the docker rmi command in the ./docker_run.sh script to remove them.