This section contains information regarding
the features and updates of the Vitis AI Library
3.5 release.
Key Features And Enhancements
This Vitis AI Library release includes the following key features and enhancements:
- New Board Support
- AMD Versal™ VEK280, V70 evaluation board are supported in this release.
- ONNX Runtime Enhanced
- ONNX Runtime Vitis AI Execution Provider (Vitis AI EP) is provided to hardware-accelerated AI inference with DPUs. It supports both C++ and Python API in this release.
- New Model Libraries
- Following are the new model libraries supported in this release.
- New Model Support
-
- Added four new PyTorch models
- Added one new TensorFlow2 model
Changes
For Vitis AI 3.5, only the VEK280, V70 evaluation board are supported and verified. For Zynq UltraScale+ MPSoC and VCK190, VCK5000-Prod boards, refer to Vitis AI 3.0.
Compatibility
The Vitis AI Library 3.5 is tested with the following images.
- amd-vek280-dpu-v2023.1-v3.5.0.img.gz
Device Support
The following platforms and evaluation boards (EVB) are supported by the Vitis AI Library 3.5.
| Platform | EVB | Version |
|---|---|---|
| Versal Edge series VE2802 | AMD VEK280 | ES1 |
| Accelerator Cards |
|---|
| Versal AI Core series V70 Data Center development kit |
Limitations
- Zynq UltraScale+ MPSoC and VCK190 are not updated in Vitis AI 3.5 and will be updated in the next release.
- VCK5000-Prod is no longer supported. Use Vitis AI 3.0 if using VCK5000-Production cards.
- Alveo U50 and U280 cards are no longer supported. Use Vitis AI 1.4.x if using these cards.
- Due to limitations of the Docker environment, MultiTask demos cannot run in DRM mode for Data Center targets.