Learn to train a neural network model for recognizing MNIST handwritten digits in this 41-minute hands-on lab video. Master PyTorch fundamentals including dataset handling, one-hot encoding of labels, and proper data splitting between training, validation, and testing sets. Explore the implementation of DataLoaders for mini-batch processing, construct effective training loops with loss data collection and validation, and visualize learning progress through loss curves. Gain practical experience in saving and loading models for inference, and understand the differences between incremental and full-batch updates. Access the provided Colab notebook to follow along with the step-by-step implementation, from initial setup through model deployment. Enhance your deep learning skills by implementing gradient descent and mini-batch updates in a real-world computer vision application.
Training MNIST Perceptron with PyTorch - Lab 3.4
Overview
Syllabus
- Recap
- PyTorch Datasets, One-Hot Encoding of Labels
- Data Split Training, Validation, Testing
- PyTorch DataLoaders and Mini Batches
- The Training Loop for a Neural Network
- Improved Training Loop Loss Data Collection & Validation
- Running the Training Loop
- Plotting the Loss Curves
- Saving and Loading PyTorch Models for Inference
- Incremental and Full-Batch Updates
Taught by
Donato Capitella
Related Courses
-
PyTorch for Deep Learning & Machine Learning – Full Course
-
Introduction to Neural Networks and PyTorch
-
Loss Function and Gradient Descent in Neural Network Training - Part 3.1
-
Foundations and Core Concepts of PyTorch
-
Create a Large Language Model from Scratch with Python – Tutorial
-
Implementing and Training a Neural Network with PyTorch
