Start our project on your local machine!

ATTENTION: Since we did lots of operations on the GCP(for example, downloading the pre-trained model from cloud storage), you also need to create your own GCP account and replace the project id, bucket name etc.

1. Download the overall project files

1. Create your directory in your local machine.

2. Folk our repository / Create the git file and clone our repository

   git clone https://github.com/ZhuMuMu0216/MLOps.git
   

2. Create venv

1. Create the virtual environment

   python -m venv venv       # create environment
   
   source venv/bin/activate  # Linux/macOS
   .\venv\Scripts\activate   # Windows
   
   pip install .             # install all the packages
   

3. Download the dataset

1. You should enter the MLOps directory.

2. Run the below command in your terminal.

   dvc pull
   

4. Run the code in docker

1. You should enter the MLOps directory.

2. Run the below command in your terminal, and you will build the docker image based on my dockerfile.

   ```bash docker build -t train_image -f dockerfiles/train.dockerfile .

3. Check your built docker image.

   docker images
   

   You will get below result and we can run the train_image now.

   REPOSITORY	TAG	IMAGE ID	CREATED	SIZE
   train_image	latest	311535037766	8 minutes ago	6.24GB

4. As we didn't COPY the data into our docker image, we dynamically mount the Host's data Directory to the Container's /data.

   '''In Linux system'''
   docker run -v $(pwd)/data:/data -it train_image
   
   '''In Windows Shell'''
   docker run -v ${PWD}/data:/data -it train_image
   

5. Deploy the API

   docker build -t api-service -f api.dockerfile .
   docker run -d -p 8080:8080 api-service
   

   You can access your API at http://localhost:8080

---