search

gofynd / mildnet

Visual Similarity research at Fynd. Contains code to reproduce 2 of our research papers.
https://arxiv.org/abs/1903.00905
Apache License 2.0
84 stars 33 forks source link

Why did you use count//batch_size*batch_size? #3

Open samehraban opened 5 years ago

samehraban commented 5 years ago

In trainer.datagen.Iterator class constructor, you counted lines in csv file and then set size of dataset in this way:

count = count//batch_size*batch_size

This way you make sure that count is divisible by batch_size, but is it necessary?