This repo has two MNIST notebooks.
-
MNIST-PySGD.ipynb implements SGD as built-in PyTorch Module
-
MNIST-MySGD.ipynb has SGD function in the notebook.
Training process can be run on cpu and is iterative in the sense that it checks if the trained model is already available then it starts from there.
The best weights are saved and loaded based on loss.
Please submit your feedback to Nagender Aneja. Please write an email (naneja@gmail.com)