/snail

A PyTorch implementation of the blocks from the _A Simple Neural Attentive Meta-Learner_ paper

Primary LanguagePythonMIT LicenseMIT

A Simple Neural Attentive Meta-Learner implementation in PyTorch

A PyTorch implementation of the SNAIL building blocks.

This module implements the three blocks in A Simple Neural Attentive Meta-Learner by Mishra et al.

The three building blocks are the following:

  • A dense block, built with causal convolutions.
  • A TC Block, built with a stack of dense blocks.
  • An attention block, similar to the attention mechanism described by Vaswani et al (2017).