The Vitis AI 2.0 release uses containers to distribute the AI software. The release consists of the following components.
- Tools container
- Runtime package for Zynq UltraScale+ MPSoC
- Public GitHub for examples (https://github.com/Xilinx/Vitis-AI)
- Vitis AI Model Zoo (https://github.com/Xilinx/Vitis-AI/tree/master/models/AI-Model-Zoo)
Tools Container
The tools container consists of the following:
- Containers distributed through Docker Hub: https://hub.docker.com/r/xilinx/vitis-ai/tags
- Unified compiler flow includes:
- Compiler flow for DPUCZDX8G (Edge)
- Compiler flow for DPUCAHX8H (Cloud)
- Compiler flow for DPUCVDX8G (Edge)
- Compiler flow for DPUCVDX8H (Cloud)
- Compiler flow for DPUCADF8H (Cloud)
- Pre-built conda environment to run frameworks:
-
conda activate vitis-ai-caffe
for Caffe-based flows -
conda activate vitis-ai-tensorflow
for TensorFlow-based flows -
conda activate vitis-ai-tensorflow2
for TensorFlow2-based flows -
conda activate vitis-ai-pytorch
for PyTorch-based flows
-
- Alveo Runtime tools
Runtime Packages for Zynq UltraScale+ MPSoC Devices
The runtime packages are located at https://github.com/Xilinx/Vitis-AI/tree/master/setup/mpsoc/VART. It contains the following items:
- PetaLinux SDK and Cross compiler tool chain
- Vitis AI board packages based on the 2021.2 release, including the Vitis AI new generation runtime VART.
Models and overlaybins are located at https://github.com/Xilinx/Vitis-AI. You can also find the following items here:
- All public pre-trained models
- Overlays for Zynq UltraScale+ MPSoCs and Alveo accelerator cards
- Scripts to automate the downloading and installation processes for models and overlays.