jimfleming/recurrent-entity-networks

Reproducibility on Task 14

anantzoid opened this issue · 1 comments

anantzoid commented

Hi Jim,

I'm can't seem to get a 100% accuracy on task 14 as reported on README. Here are some stats:

screen shot 2018-10-30 at 9 30 40 pm

screen shot 2018-10-30 at 9 30 51 pm

anantzoid commented

There's a reproducibility issue on Task 19 as well. Running the code for 53041 global steps reports: "Best step: 37440 with accuracy = 0.412", i.e., 0.6 error, while the repo reports 0 error.

Also, the gradient norms seems to be increasing which is unusual. Is this expected?
screen shot 2019-01-04 at 6 13 03 pm