Skip to content

Instantly share code, notes, and snippets.

View nathanin's full-sized avatar
coffee

Nathan Ing nathanin

coffee
View GitHub Profile
"""
Distributed MNIST to demonstrate slow autographing of larger models when using a MirroredStrategy
Using the VGG16 model ~ 8 minutes to start iterating on 4 GPUs
Using the 2-CNN model ~ 1 minute to start iterating on 4 GPUs
"""
import tensorflow as tf
import numpy as np