To get started with this workshop, you can refer to the odsc_workshop_data.ipynb notebook. This notebook is divided into two key sections:
-
Optimizing ImageNet with Lightning: The first part of this notebook provides a step-by-step walkthrough of the process required to use Lightning to optimize ImageNet.
-
Training ResNet Efficiently: The second part utilizes the optimized ImageNet to train ResNet models swiftly and efficiently, allowing you to leverage the benefits of the ImageNet optimization for your model training.
In the second part of this workshop, we use odsc_workshop_train.py to train ResNet18 on a subset of the data using the Lightning trainer. This part illustrates how simple it is to implement various training optimizations using Lightning, such as multi-GPU training and different precision techniques.