---

Tiny ONNC is an easy-to-use AI compiler for microcontrollers: converting popular neural network models into a function within a series of function calls of neural network libraries.

---

Tiny ONNC for CMSIS-NN r1.1.0 is out!

List of new features:

• Upgrade CMSIS-NN library to tag NN/2.0.0

• Output C source code conforms to C99 standard

• Enable to turn off preprocessing

• Accelerate operator by CMSIS-NN *_fast() and arm_convolve_HWC_q7_RGB()

• Support new operators: Add

• Ease operator constraints: AveragePool, Conv, depth-wise Conv

Getting Started

• Getting Started
  • Installation
  • Acquire Pre-trained Models
  • Register An Account
  • Compile the Model
  • Review the Compilation Results
  • Integrate with your Application
  • ONNC Forest Runtime

Application Programming Interface

• C API
  • Data Structure
    • Neural Network Model
    • Raw Tensor
  • Standard Procedures
    • Getting Neural Network Model
    • Writting Inputs
    • Model Inference
    • Reading Outputs
    • Release The Model
  • Example
• Python APIs
  • Example
  • login function
  • Project Class
  • Project.add_model Method
  • Project.compile Method
  • Project.save Method
  • Deployment Class

Tutorial

• Installation
  • Install ONNC Bench by pip
  • Install Virtual Environment Managers
• Registration
• Python APIs
  • Example
  • login function
  • Project Class
  • Project.add_model Method
  • Project.compile Method
  • Project.save Method
  • Deployment Class
• NVDLA Backend Examples
  • 1. Compile a Pytorch Resnet Model
  • 2. Compile a Keras Resnet model
  • 3. Compile a ONNX Model Zoo model
• OpenVino Backend Examples
  • Prerequisites
  • A Tensorflow Example
  • A Pytorch Example

Manuals

• Supported Frameworks
• Supported Devices