Best practices for generating a random seed to seed Pytorch?

pinocchio · August 8, 2019, 4:18pm

What I really want is to seed the dataset and dataloader. I am adapting code from:

gist.github.com

https://gist.github.com/kevinzakka/d33bf8d6c7f06a9d8c76d97a7879f5cb

data_loader.py

"""
Create train, valid, test iterators for CIFAR-10 [1].
Easily extended to MNIST, CIFAR-100 and Imagenet.

[1]: https://discuss.pytorch.org/t/feedback-on-pytorch-for-kaggle-competitions/2252/4
"""

import torch
import numpy as np

This file has been truncated. show original

utils.py

import matplotlib.pyplot as plt

label_names = [
    'airplane',
    'automobile',
    'bird',
    'cat',
    'deer',
    'dog',
    'frog',

This file has been truncated. show original

Anyone know how to seed this properly? What are the best practices for seeding things in Pytorch.

Honestly, I have no idea if there is an algorithm specific way for GPU vs CPU. I care mostly about general pytorch and make sure my code is “truly random”. Specially when it uses GPU I guess…

pinocchio · August 8, 2019, 4:18pm