Skip to content

Latest commit

 

History

History
20 lines (15 loc) · 861 Bytes

README.md

File metadata and controls

20 lines (15 loc) · 861 Bytes
Raw

Paper

Attention pooling: https://arxiv.org/pdf/1602.03609.pdf Fastformer: https://arxiv.org/pdf/2108.09084.pdf

How to run

Google Colab

In the Google Drive root folder

  • Create /Projects/MIND-FUM/ folder
  • Create /Dataset/MIND/data folder
  • Create /Dataset/MIND/embeddings folder
  • Download and extract MIND dataset to /Dataset/MIND/data folder (https://msnews.github.io/)
  • Download GloVe embeddings (840B.300d) to /Dataset/MIND/embeddings folder (https://nlp.stanford.edu/projects/glove/)
  • Upload Generator.py, Hypers.py, Models.py, Utils.py, Preprocessing.py to the /Projects/MIND-FUM/
  • Run the FUM-train.ipynb notebook to train the model in Google Colab
  • Run the FUM_code_eval.ipynb notebook to evaluate the model in Google Colab
  • Go to Google Drive and use the /Dataset/MIND/prediction.txt to submit to the competition