A collection of starter notebooks for DTC Zoomcamp Q&A challenge on Kaggle on Kaggle
In this notebook, we show how to:
- Download the data from Kaggle and unzip it
- Read the data
- Use BERT to create embeddings
- Calculate the cosine similarity between questions and answers
- Select the most similar answer as prediction
- Submit the results
See the notebook here: bert-starter.ipynb
You can run this notebook in SaturnCloud:
To use the recipe:
- Download the credential file from Kaggle
- Put the content of the file to SaturnCloud secrets, save this secret as "kaggle"
- Click on the button above to create a resource in SaturnCloud
- Verify that the kaggle secret is linked in the "secrets" tab
- Run the code and submit your predictions
- Improve the score
You can also see it as a video:
(The video was originally created for another competition, but it works for this one too)
TBA