Deep Learning Boilerplate code or Scaffolding.

A boilerplate to get started with any Deep Learning project (Vision)

Dataset:

MNIST dataset downloaded using torchvision

Experiments

Shared parameters:

batch size: 32
Image dimensions (HxW): 28x28
Model: LeNet
initial learning rate: 3e-4
epochs: 10

Dataset-split: 10% for Validation

Experiment 1

Loss
Training set 0.007
Evaluation set 0.031

plot