Vitis AI Optimizer Overview - 2.5 English

Vitis AI Optimizer User Guide (UG1333)

Document ID
Release Date
2.5 English

Vitis™ AI is a Xilinx® development kit for AI inference on Xilinx hardware platforms. Inference in machine learning is computation-intensive and requires high memory bandwidth to meet the low-latency and high-throughput requirements of various applications.

The Vitis AI optimizer provides the ability to optimize neural network models. Currently, the Vitis AI optimizer includes only one tool called the pruner. The Vitis AI pruner prunes redundant kernels in neural networks thereby reducing the overall computational cost for inference. The pruned models produced by the Vitis AI pruner are then quantized by the Vitis AI quantizer and deployed to a Xilinx FPGA, SoC, or ACAP devices. For more information on the Vitis AI quantizer and deployment, see the Vitis AI User Guide (UG1414).

Figure 1. VAI Optimizer

The Vitis AI pruner supports four deep learning frameworks. The frameworks and their corresponding tool names (_p_ denotes pruning) are listed in the following table:

Table 1. Vitis AI Pruner Frameworks and Tool Names
Framework Tool Name
TensorFlow vai_p_tensorflow (TF1.15), vai_p_tensorflow2 (TF2.x)
PyTorch vai_p_pytorch

The Vitis AI optimizer requires a commercial license. Contact to obtain access to the Vitis AI optimizer installation package and license.