Text_Summarizer_PyTorch

This project aims to build a text summarizer with frontend using PyTorch and Hugging Face.

Instructions to use

• Pull the image from DockerHub:

  docker pull antonbeloval08/text-summarizer

OR

• Build the image yourself:

  git clone https://github.com/Commit2Cosmos/Text_Summarizer_PyTorch.git
  docker build --no-cache -t text-summarizer .

• Create a container:

  docker run --name text-sumarizer-test -p 8000:8000 text-summarizer

• Visit the webapp: http://0.0.0.0:8000

• To train: Click on the "Training" section, press "Try it out" and press "Execute"

• To summarize: Click on the "Inference" section, press "Try it out", enter the text you want summarized and press "Execute"

-------------------------- DEV NOTES --------------------------

TODO

• Add ability to control parameters in params.json in the web app

• Change training + evaluation components to work with multiple datasets (dynamic saving file paths etc)

• Resolve Some non-default generation parameters are set in the model config. These should go into a GenerationConfig file

• Raise issue of the incorrect warning about that model needs to be trained because of newly initialised encoding layers

Milestones

• Choose and download the pre-trained model and dataset for transfer learning
• Build pipelines (listed below)
• Check for transfer learning or fine-tuning (Untrained layers are already provided, so just train them -> update training pipeline)
• Model packaging (serialisation, containerisation) -> Docker
• Choose deployment strategy (cloud or local) and interaction type (API, webapp, cli, embedded systems)
• Train the model with VertexAI (separate data and model -> don't use the container) and upload trained weights

(Optional):

• Build + deploy custom webapp
• Support for multiple datasets
• Add output text size control feature
• Add context area for user defined personalisations
• Add support for pdf (and other) files (multimodality)

Pipelines

• Data Ingestion
• Data Validation
• Data Transformation + Feature Engineering
• Model Training
• Model Evaluation
• Model Inference

Workflow (Files to update)

See architecture file for detailed breakdown of the project's architecture and what each file does.

• logging.py

• pyproject.toml

• params.json

• config.json

• src/entity

• src/config

• src/components

• src/pipeline

• main.py

• app.py