Open, Modular, Deep Learning Accelerator

Clone this repo:
  1. 4319417 adapt chisel impl to new VTA ISA (#24) by Luis Vega · 29 hours ago main
  2. b61fe13 [CI] Fix build on main (#25) by Tianqi Chen · 2 days ago
  3. f11ca65 Use TVMArrayCopyFromBytes API to replace VTA internal function VTABufferCPUPtr (#23) by Hua Jiang · 4 weeks ago
  4. f187a39 Make VTA C++ Deploy Compatible with latest TVM. (#22) by Hua Jiang · 5 weeks ago
  5. 5bd9c6a [Hardware][OpenCL] Intelfocl support (#9) by ZHANG Hao · 4 months ago

VTA Hardware Design Stack

Build Status

VTA (versatile tensor accelerator) is an open-source deep learning accelerator complemented with an end-to-end TVM-based compiler stack.

The key features of VTA include:

  • Generic, modular, open-source hardware
    • Streamlined workflow to deploy to FPGAs.
    • Simulator support to prototype compilation passes on regular workstations.
  • Driver and JIT runtime for both simulator and FPGA hardware back-end.
  • End-to-end TVM stack integration
    • Direct optimization and deployment of models from deep learning frameworks via TVM.
    • Customized and extensible TVM compiler back-end.
    • Flexible RPC support to ease deployment, and program FPGAs with the convenience of Python.