T5 Summarisation Using Pytorch Lightning, DVC, DagsHub and HuggingFace Spaces

Here you will find the code for the project, but also the data, models, pipelines and experiments. This means that the project is easily reproducible on any machine, but also that you can contribute data, models, and code to it.

Have a great idea for how to improve the model? Want to add data and metrics to make it more explainable/fair? We'd love to get your help.

Blog: https://dagshub.com/blog/machine-summarization-an-open-data-science-project/

To use and run the DVC pipeline install the t5s package

pip install t5s

Firstly we need to clone the repo containing the code so we can do that using:

t5s clone 

We would then have to create the required directories to run the pipeline

t5s dirs

Now to define the parameters for the run we have to run:

t5s start [-h] [-d DATASET] [-s SPLIT] [-n NAME] [-mt MODEL_TYPE]
                 [-m MODEL_NAME] [-e EPOCHS] [-lr LEARNING_RATE]
                 [-b BATCH_SIZE]

Then we need to pull the models from DVC

t5s pull

Now to run the training pipeline we can run:

t5s run

Before pushing make sure that the DVC remote is setup correctly:

dvc remote modify origin url https://dagshub.com/{user_name}/summarization.dvc
dvc remote modify origin --local auth basic
dvc remote modify origin --local user {user_name}
dvc remote modify origin --local password {your_token}

Finally to push the model to DVC

t5s push

To push this model to HuggingFace Hub for inference you can run:

t5s upload

Next if we would like to test the model and visualise the results we can run:

t5s visualize

And this would create a streamlit app for testing