RUHMI (Robust Unified Heterogeneous Model Integration) provides an AI model compiler workflow for Renesas RZ/G3E.
This repository includes installation assets, model compilation scripts, and application examples.
RUHMI Framework provides tools to compile machine learning models into deployment artifacts compatible with RZ/G3E.
The AI compiler stack is powered by EdgeCortix® MERA™.
This repository provides the guidance for how to use RUHMI AI model compiler. Also provides some application examples with the the guidance for how to run on RZ/G3E-EVKIT.
- Renesas MPU RZ/G3E
[NOTE]
The evaluation board, RZ/G3E-EVKIT, currently on sale includes V1 device which does not support Ethos operation using LPDDR4/4X. Production is underway for the board with V2 sample, with stock expected to arrive at the end of June. To identify which sample is installed on an board, check the S.LOT No. printed on the board box.
Less than 0000300000: V1 sample installed
0000300000 or later: V2 sample installed
Lot No. placement
- TensorFlow Lite/INT8
[NOTE] The current version is the 1st release of RHUMI for RZ/G3E. The supported AI model is limited within 16MB.
- Build the envronment and install RUHMI AI model compiler
- Cimpilae tha target model
This repository provide the simpliped expanation. You can also refer to Renesas Rz/G3E NPU(Ethos0U55) SUpport.
application_examples/: runnable sample apps and app-specific docs for RZ/G3Edocs/: supporting documentation and documentation assetsinstall/: MERA/RUHMI install artifacts (wheel files, shared libraries) and setup guidescripts/: Docker build environment and model data generation scriptrequirements-host.txt: host-side Python dependency listREADME.md: repository entry pointLICENSE.md: license terms
RZ/G3E Ethos Support Package.
Renesas RZ/G3E NPU (Ethos-U55) Support
Evaluation Board Kit for RZ/G3E MPU, RZ/G3E-EVKIT
See LICENSE.md.
