The previous chapter introduced the principles and implementation for a simple deep learning model, including multi-layer perceptrons. In this chapter we will cover various key components of deep learning computation, such as model construction, parameter access and initialization, custom layers, and reading, storing, and using GPUs. Throughout this chapter, you will gain important insights into model implementation and computation details, which gives readers a solid foundation for implementing more complex models in the following chapters.
- Layers and Blocks
- Parameter Management
- Deferred Initialization
- Custom Layers
- File I/O
- GPU's