INT8 quantization in PyTorch from scratch — scale factors, zero points, and integer arithmetic without QuantStub, to understand exactly what quantized layers do under the hood.

In this article, we will discuss how values are passed between layers post quantization in PyTorch. We will also discuss why floating point operations are slower than integer operations.