After cloning the repository:

• Navigate to the project directory and create a virtual environment for the project: python -m venv trocr_torchserve
• Activate the virtual environment:
  • Windows: .\trocr_torchserve\Scripts\activate
  • macOS/Linux: source trocr_torchserve/bin/activate
• Install all required packages: pip install -r requirements.txt
• Run the setup.sh script.

Now, run the following scripts to archive the model and launch torchserve respectively:

• archive_model.sh <model_name> <version_number> - script that will create the torchserve .mar file for model serving. Example usage: archive_model.sh trocr_base 1.0
• start_server.sh - starts the torchserve server using configurations specified in config.properties

Once the server is running, you can run the client.py script which will send images from the Teklia/IAM-line Hugging Face dataset to the model server for inference. The usage is: python cilent.py --ids <csv_list_of_ids>. For example python client.py --ids 1,5,9,13 sends images with the sample ids 1, 5, 9, and 13 from the dataset to the model server for inference.

• Note, if running the client.py script, each image will be displayed in a new window. Once each window in a batch is closed, the data will be sent to the server.