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
1.49k stars 633 forks source link

How to run a compiled model on the hardware (VCK190) ? #1479

Open MeltedHyperion07 opened 1 month ago

MeltedHyperion07 commented 1 month ago

So we are done with quantization and compilation of our custom model (landmark regression) based on HRNet. (Small issue with compilation : #1478 ) Now we need to run it on the hardware. We looked into the documentation and it seems we need to use the Vitis AI runtime and library but we couldnt find any proper documentation for how to use it. (I believe this is what we used for running the resnet50 example. For the pytorch resnet18 example we had these files for executing the xmodel as well: image image image

we used to build these files and then execute the xmodel using the builds

[Target] $ ./test_jpeg_classification resnet18_pt ~/Vitis-AI/examples/vai_library/samples/classification/images/002.jpg
[Target] $ ./test_video_classification resnet18_pt ~/Vitis-AI/examples/vai_library/seg_and_pose_detect/pose_960_540.avi -t 8

can someone help us figure out how to make these and what the flow for this should be? @hanxue @eemars @quentonh @woinck