search

CSL-KU / firesim-nvdla

FireSim-NVDLA: NVIDIA Deep Learning Accelerator (NVDLA) Integrated with RISC-V Rocket Chip SoC Running on the Amazon FPGA Cloud
Other
161 stars 31 forks source link

about unsupported layers in YOLO #7

Closed cnjsdfcy closed 4 years ago

cnjsdfcy commented 4 years ago

Hi Farzad,

  1. How do you compose the loadable file for YOLO? NVDLA compiler will report errors when encounter unsupported layers.
  2. After you get the loadable file, how do you dispatch these unsupported layers to RISC-V processor? I checked the KMD src, it seems that you did not modify the scheduler.

Thanks, Caiyu

farzadfch commented 4 years ago

This is done by @prasshantg. Darknet already has the implementation for the YOLO in C code, and as far as I know, we are using the Darknet implementation for the unsupported layers and run them on the RISC-V processor.

cnjsdfcy commented 4 years ago

Thank you for your response, Farzad. I found this issue, seems that others also wonder how three separated nvdla loadable files were generated.

cnjsdfcy commented 4 years ago

close thanks to Prashant's answer.