Motivation

Inspired by fast.ai course, we want to build our own Deep Learning framework.

Therefore, we can dive into Deep Learning by implementing papers from scratch.

We will not aim to replace fast.ai library. Instead, we want to boost our deep learning knowledge and contribute to fast.ai library if we can.

The main goal of this project is to:

  • Provide a simple framework for fast prototyping of Deep Learning papers.

  • Learn from other frameworks and continuously refactor the framework.

  • Detailed understanding of:

    • Parameters initialization
    • Normalization techniques: BatchNorm, InstanceNorm, GroupNorm …
    • Visualize activations / gradients histograms.