deepstream_4.x_apps icon indicating copy to clipboard operation
deepstream_4.x_apps copied to clipboard

Published 20 hours ago • verified publisher icon NVIDIA-AI-IOT

FP16 fix for Deepstream 4.0.2

Open kingardor opened this issue 4 years ago • 2 comments

Deepstream 4.0.2 uses TensorRT6 and the fp16 fix seems to be built for TensorRT5. Is there a way I could compile the fix myself?

kingardor avatar Mar 09 '20 17:03 kingardor

Using this fix (7. FP16 model issue), I was able to build a FP16 engine for a MaskRCNN model on Deepstream 4.0.2 with TensorRT 6.0.1.

TheophileBlard avatar Mar 11 '20 15:03 TheophileBlard

Thanks @TheophileBlard Will give it a shot.

kingardor avatar Mar 11 '20 17:03 kingardor