Giters

bclarkson-code / Tricycle

Deep learning framework completely from scratch in python + numpy

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Optimise embedding

bclarkson-code opened this issue · comments

commented

Currently, the embedding layer uses more memory and is slower than I would expect. It should be optimised

commented

I ran a benchmark on the new code and the speedup is huge! ~8.5x faster

┏━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━━━┳━━━━━━━━━┳━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┓
┃ Benchmark ┃ Min ┃ Max ┃ Mean ┃ Min (+) ┃ Max (+) ┃ Mean (+) ┃
┡━━━━━━━━━━━╇━━━━━━━━━╇━━━━━━━━━╇━━━━━━━━━╇━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━┩
│ original │ 0.677 │ 0.744 │ 0.692 │ 0.080 (8.5x) │ 0.081 (9.2x) │ 0.080 (8.6x) │
└───────────┴─────────┴─────────┴─────────┴─────────────────┴─────────────────┴─────────────────┘