search

Xilinx / Vitis-AI

Vitis AI is Xilinx’s development stack for AI inference on Xilinx hardware platforms, including both edge devices and Alveo cards.
https://www.xilinx.com/ai
Apache License 2.0
1.48k stars 628 forks source link

dpu_yolov4 kernal crashing #450

Open daivikbhatia opened 3 years ago

daivikbhatia commented 3 years ago

Hi, I'm converting a yolo-v4 darknet model into tensorflow. The conversion, quantization and compilation was sucessful. I got the elf file and the kernal info. The kernal name in kernal info is "dpu_yolov4" but as soon as I put this kernal name in n2cube.dpuLoadKernal("dpu_yolov4"), the jupyter kernal of pynq 2.5 image crashes. Also I tried same darknet model for caffe with kernal name "yolov4_voc" and everything ran smoothly. Can you help me resolve it? the deployment reference I took was from here. https://github.com/Xilinx/Vitis-AI/blob/1.1/demo/DNNDK/tf_yolov3_voc_py/tf_yolov3_voc.py

zackdilan commented 3 years ago

Can you show the Jupyter kernel log ??

daivikbhatia commented 3 years ago

Here is a screenshot of kernal log

Screenshot (84)

zackdilan commented 3 years ago

Hi,

Thanks, just wanted to know! Have you already troubleshot the error showing in the log?

daivikbhatia commented 3 years ago

Hi, No I'm still stuck here. Do you have any solution?

zackdilan commented 3 years ago

I m sorry I cannot help you for the time being(I should find some time to set up DPU-PYNQ). But just make sure you can run the inference on the DPU image(not pynq) using regular dnndk runtime from Xilinx(To sanity-check the dpu kernel you got). There are demos on using dnndk APIs to create the inference application.