This tutorial demonstrates how to apply INT8 quantization to the Natural Language Processing model BERT, using the Post-Training Quantization API. The HuggingFace BERT PyTorch model, fine-tuned for Microsoft Research Paraphrase Corpus (MRPC) task is used. The code of this tutorial is designed to be extendable to custom models and datasets.
The tutorial consists of the following steps:
- Downloading and preparing the MRPC model and a dataset.
- Defining data loading functionality.
- Running optimization pipeline.
- Comparing F1 score of the original and quantized models.
- Comparing performance of the original and quantized models.
If you have not installed all required dependencies, follow the Installation Guide.