AXera Pulsar2 Toolchain Instructions Manual

copyright

• 1. Copyright Notice

Pulsar2 Toolchain

• 1. Documentation modifications and release notes
  • 1.1. Modify history
• 2. Pulsar2 Toolchain overview
  • 2.1. Introduction
  • 2.2. Introduction to Virtual NPU
  • 2.3. Guide to the content of subsequent chapters
• 3. Development environment preparation
  • 3.1. Install the Docker development environment
  • 3.2. Install Pulsar2 toolchain
• 4. Quick Start(AX650)
  • 4.1. Pulsar2 toolchain command description
  • 4.2. Model compilation configuration file description
  • 4.3. Compile and execute
  • 4.4. Simulation Run
  • 4.5. Development board running
• 5. Quick Start(AX620E)
  • 5.1. Pulsar2 toolchain command description
  • 5.2. Model compilation configuration file description
  • 5.3. Compile and execute
  • 5.4. Simulation Run
  • 5.5. Development board running
• 6. Quick Start(AX615)
  • 6.1. Pulsar2 toolchain command description
  • 6.2. Model compilation configuration file description
  • 6.3. Compile and execute
  • 6.4. Simulation Run
  • 6.5. Development board running
• 7. Quick Start(M57)
  • 7.1. Pulsar2 toolchain command description
  • 7.2. Model compilation configuration file description
  • 7.3. Compile and execute
  • 7.4. Simulation Run
• 8. Quick Start(AX637)
  • 8.1. Pulsar2 toolchain command description
  • 8.2. Model compilation configuration file description
  • 8.3. Compile and execute
  • 8.4. Simulation Run
  • 8.5. Development board running
• 9. Advanced Guide to Model Conversion
  • 9.1. Overview
  • 9.2. Detailed explanation of model compilation
  • 9.3. Detailed explanation of multi-core compilation
  • 9.4. Detailed explanation of multi-batch compilation
  • 9.5. Detailed explanation layer by layer
  • 9.6. Detailed explanation of loading custom data sets
  • 9.7. Multi-input model configuration quantitative data set
  • 9.8. Multi-input model configuration quantization data set (NumpyObject)
  • 9.9. Detailed explanation of mixed precision quantization
  • 9.10. Enter size modification
  • 9.11. Configure model additional input dimensions
  • 9.12. Operator attribute modification
  • 9.13. Constant data modification
  • 9.14. Set separate compilation options for subgraphs
  • 9.15. Transformer model configuration details
  • 9.16. Quantized ONNX model import
  • 9.17. Color space conversion configuration
  • 9.18. Detailed explanation of advanced quantitative strategy configuration
• 10. Advanced Guide to Simulation Running
  • 10.1. Overview
  • 10.2. Detailed explanation of simulation operation
  • 10.3. Simulation run example
• 11. Advanced Guide to Model Deployment
  • 11.1. Overview
  • 11.2. Other Examples
• 12. Configuration file details
  • 12.1. Profile overview
  • 12.2. Complete json configuration reference
  • 12.3. Quantitative parameter description
  • 12.4. Quantitative precision analysis parameter description
  • 12.5. Preprocessing and postprocessing parameter description
  • 12.6. proto configuration definition
• 13. Instructions for using model evaluation tools
  • 13.1. Parameter Description
  • 13.2. Usage example
• 14. Functional Safety

Other Tools

• 1. Tool usage instructions of caffe2onnx
  • 1.1. Convert Caffe model to ONNX model
  • 1.2. Validate the converted ONNX model
• 2. QAT 4W4F support
• 3. Polygraphy tool overview
  • 3.1. Common parameters
  • 3.2. Quick start
• 4. NPU trace tool instructions
  • 4.1. NPU trace overview
  • 4.2. Get trace.json
  • 4.3. Trace features
  • 4.4. Common issue analysis

appendix

• 1. NPU Operators support list (AX650)
• 2. NPU Operators support list(AX620E)
• 3. NPU Operators support list (AX615)
• 4. NPU Operators support list(M57)
• 5. NPU Operators support list (AX637)
• 6. Accuracy Tuning Suggestions
• 7. LLM Build (Compile)
• 8. Model conversion examples
• 9. Common configuration examples