Listen, Attend and Spell: Pyramidal Length
Beginner
The problem, from first principles
Listen, Attend and Spell (Chan et al., 2016) reads thousands of speech frames. A pyramidal encoder halves the time resolution at each layer, so the attention decoder works over a much shorter sequence.
From theory to code
Implement pyramid_length(T, layers), the length after the pyramidal reduction.
Constraints
- Round up at each halving.
Hints
Hint 1
Divide the length by two to the power of the layer count, rounding up.
Keywords
research-paperssequence-modelsspeechattention
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.