Dockerfile now loads models to local folder. Can use custom output folder. requrirements for GPU-enabled summarisation now in separate file to hopefully avoid HF space issues.

Upgraded Gradio version. Changed large model to Phi 3 128k. Now uses Llama cpp Python

Added a stop processing button and timeout for Mistral model

Couple of changes to requirements

Changed small and large models. Large model should now work for summarisation

Initialise app with basic functionality