search

better / convoys

Implementation of statistical models to analyze time lagged conversions
https://better.engineering/convoys/
MIT License
254 stars 42 forks source link

use constants instead of placeholders for inputs #22

Closed erikbern closed 6 years ago

erikbern commented 6 years ago

realized it's really slow to feed 400MB of data into tensorflow on each step

this speeds it up like 10x

coveralls commented 6 years ago

Coverage Status

Coverage decreased (-0.1%) to 85.197% when pulling 75520da5aefc6474f72c829f9499b8276fe94c44 on tf-constants into 169b02f2ccabd1448d08804c0b8ca60d052c830a on master.

coveralls commented 6 years ago

Coverage Status

Coverage decreased (-0.2%) to 85.149% when pulling 04f7636cff532096e2da169784b39b0d656f86ca on tf-constants into 169b02f2ccabd1448d08804c0b8ca60d052c830a on master.

erikbern commented 6 years ago

weird, on my mega test, it takes the number of steps from 1400 up to 6000, but it still runs in half the time. not sure what's up

erikbern commented 6 years ago

really need to figure out how to use minibatches