The ModusToolbox™ ML Middleware Library is a set of helper functions to handle the Machine Learning models and capability to receive stream validation data.
Please refer to the README.md for a complete description of the ModusToolbox™ ML Middleware Library.
This release of the ModusToolbox™ ML Middleware library includes support for the following functionality:
- utils - functions to help handling the inputs/outputs of a machine learning model
- model - functions to initialize and inference a ML model
- stream - functions to stream validation data
Support floating-point and fixed-point variants:
- 32-bit floating-point
- 16-bit fixed-point input
- 8-bit fixed-point input
- 16-bit fixed-point weight
- 8-bit fixed-point weight
The MTB-ML Middleware library supports Windows, Linux and Mac OS.
- Added mtb_ml_utils_model_quantize & mtb_ml_utils_convert_tflm_flt_to_int8 functions
- Support for TensorFlow Lite Micro
- Initial release
This version of the was validated for compatibility with the following Software and Tools:
| Software and Tools | Version |
|---|---|
| ML Inference | 2.0.0 |
| ML TFLiteU | 2.0.0 |
| ML Core tools | 2.0.0 |
| GCC Compiler | 10.3-2021.07 |
| ARM Compiler 6 | 6.16 |
| IAR Compiler | 9.30.1 |
| CyBridge | 3.4.0 |
Minimum required ModusToolbox™ Software Environment: v3.0
Use the following links for more information, as needed:
- ModusToolbox™ Machine Learning Design Support
- Cypress Semiconductor, an Infineon Technologies Company
- Cypress Semiconductor GitHub
- ModusToolbox
© 2021-2022, Cypress Semiconductor Corporation (an Infineon company) or an affiliate of Cypress Semiconductor Corporation.