The Vollo-Trees SDK is designed for low latency streaming inference of decision tree models on FPGA platforms.
Note
This is the Vollo product for accelerating decision trees, the parent product for accelerating neural networks can be found at https://github.com/MyrtleSoftware/vollo-sdk
The latest SDK is available for download from https://github.com/MyrtleSoftware/vollo-trees-sdk/releases.
Download the vollo-trees-sdk-<version>.run self-extractable archive and execute it
to extract the Vollo Trees SDK contents to the current
directory:
chmod +x vollo-trees-sdk-<version>.run
./vollo-sdk-<version>.runThe user guide is available online at https://vollo-trees.myrtle.ai/.
It is also included locally as markdown at docs/mdbook/src or in the release as html:
open docs/html/index.htmlThe product architecture is shown in the diagram below:
Vollo Trees provides a C API to the user, running on the system host CPU. Vollo Trees targets FPGAs to provide low latency inference. The FPGA images (bitstreams) are prebuilt and included in the product.
Vollo Trees consists of the following elements:
-
Vollo Accelerator Bitstream. Programming file for the FPGA on the PCIe accelerator card. See the Accelerator Setup section for instructions on programming your FPGA with the bitstream.
-
Vollo Trees Compiler. Compiles decision tree models defined in ONNX to Vollo programs.
-
Vollo Runtime. The runtime library for Vollo. It sets up the the Vollo Trees accelerator with a program and provides an asynchronous inference interface for handling input and output for the accelerated model.
The Vollo Trees compiler is available to use without an accelerator card or license. Along with compiling decision tree models for the Vollo Trees accelerator, it also provides a method for estimating the performance of a model.
Refer to the Vollo Trees Compiler section in the user guide and its example walkthrough to get started with the compiler.