Documentation Overview and Installation

You can find install guidelines in installation.md.

These documents offer a comprehensive description of Intel Extension for Transformers:

• architecture.md: Provides an overview of Intel Extension for Transformers, covering various architectural levels.
• installation.md: Contains installation guidelines.

Optimization and Inference Documentation

Optimization Documentation:

• quantization.md: Covers quantization techniques.
• pruning.md: Details the pruning process.
• autodistillation.md and distillation.md: Discuss distillation methods.
• export.md: Explains PyTorch to ONNX, including the int8 model.
• get_started.md: Guides you through using the optimization API and sparse inference.

We also provide examples in examples.md. P.S., smoothquant, weight-only quantization, and cpp inference are not in these documents. SmoothQuant: smoothquant.md, weight-only quantization: weightonlyquant.md.

API Documentation:

For comprehensive API guidance, visit the API doc, generated using api_doc and build_docs. You can disregard the two mentioned folders.

Examples:

Practical examples can be found in examples.md. Note that smoothquant, weight-only quantization, and cpp inference are not covered in these documents. For specific information on these topics:

• smoothquant.md: SmoothQuant
• weightonlyquant.md: Weight-Only Quantization

Tutorials:

Explore various tutorials for different tasks in tutorials/pytorch/.

publication.md lists the publications related to this project, some of them can be found in pubs/. release.md provides links to all releases. License and legal information can be found in legal.md

Other Functionalities Documentation

Utilities:

• data_augmentation.md: Describes NLP dataset augmentation.
• benchmark.md: Explains how to measure model performance.
• metrics.md: Defines the metrics used for model measurement.
• To measure the status of the tuning model, refer to objectives.md.
• pipeline.md: Simplifies the process of using any model from the Hub for inference on tasks.

Contribution and Legal Documentation

You are invited to contribute your code. Kindly adhere to the guidelines specified in contributions.md and maintain a positive demeanor as outlined in code_of_conduct.md. For component-specific approvers, refer to component_owner.md.

Publication: publication.md lists publications related to this project, some of which can be found in pubs/.

Release Information: Access links to all releases in release.md.

Legal Information: For license and legal details, please consult legal.md.