Provides an open-source machine learning model compiler and hardware generator that creates custom inference accelerators for edge FPGAs. This enables rapid deployment of optimized ML models on resource-constrained devices, improving performance and efficiency in edge computing applications.
Funding
$150K raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Product
Problem
Deploying machine learning (ML) models to edge devices is challenging due to the limited resources and power constraints of these devices. Existing solutions often require significant manual optimization and hardware expertise, slowing down deployment and increasing costs.
Solution
Tensil provides an open-source ML model compiler and hardware generator that automates the creation of custom inference accelerators for edge field-programmable gate arrays (FPGAs). The compiler optimizes ML models for specific FPGA architectures, while the hardware generator produces synthesizable RTL (register-transfer level) code for the accelerator. This allows developers to rapidly deploy optimized ML models on resource-constrained edge devices, improving performance and energy efficiency.
Target Audience
The primary users are developers and engineers working on edge computing applications who need to deploy ML models on FPGAs, including those in robotics, IoT, and embedded systems.
Features
- Open-source ML model compiler and hardware generator
- Automated creation of custom inference accelerators for edge FPGAs
- Supports ONNX and TensorFlow frozen graphs as input model formats
- Generates synthesizable Verilog RTL code for the accelerator
- Includes a bit-accurate emulator for functional verification
- Provides tutorials and documentation for various FPGA development platforms (e.g., PYNQ Z1, Ultra96, ZCU104)
- Docker container for easy setup and deployment