Vectorblox AI

Overview

Development Flow

Documentation

Getting Started

Deployment Options

The VectorBlox Accelerator SDK contains different tools that compile a neural network description from TensorFlow and ONNX into a Binary Large Object (BLOB) that is stored in flash and loaded into the DDR memory during execution.

• Model Optimization - converts a trained network to a common intermediate representation (IR) from various training frameworks (TensorFlow, PyTorch etc.) and cleans the network for inference. Several components used during training, that slow down calculations, e.g. batching operators and fuse layers are removed at this step.
• Quantization - converts optimized networks from 32-bit floating point precision (FP32) to INT8 representation. Quantization enables networks to be represented using less memory with minimal loss in accuracy.
• Calibration - involves adjusting activations and weights of a model represented in INT8 precision.
• Runtime Generation - creates a Binary Large Object (BLOB), that is written into the SPI flash of the hardware development platform and is loaded into the DDR memory during runtime

 Step 1: Prepare your trained model

Use Python scripts provided in the SDK to convert your trained model into an optimized INT8 representation called a Binary Large Object (BLOB). Run the BLOB through the VectorBlox Simulator to verify accuracy of the network and to ensure successful conversion of the network.

 Step 2: Prepare your hardware

The PolarFire Video Kit is configured to run as an AI enabled Smart Camera. The SDK includes a pre-compiled kit image bitstream.  Write the bistream into the PolarFire FPGA using the FlashPro programmer included with the kit. Write the BLOB, generated from Step 1, into the SPI flash of the kit.

Step 3: Write your embedded code

Use the provided embedded code in C/C++ based SoftConsole IDE and generate and program the hex file. Connect the video kit to a HDMI monitor and turn it on. Modify the embedded code to load and run many CNN BLOBs, switch CNNs dynamically on the fly or load CNNs sequentially for simultaneous inferencing.