Machine Learning

SIREN: Implicit Neural Representations with Sine

Machine learning is just a function takes encoded representation of input $x$ and maps it to encoded representation of output $y$. Does the way we choose to encode the input has an impact on the model? Can we represent an image better than a

Machine Learning

10x faster RoBERTa tokenizer with Custom Tokens support

For my Question Answering Kaggle competition, I wanted to experiment replacing the BERT model with RoBERTa. This means I need to reencode and retokenize the entire Natural Questions dataset into TFRecords. This process was already taking hours with the WordPiece tokenizer used for the