ceccocats / tkDNN

Deep neural network library and toolkit to do high performace inference on NVIDIA jetson platforms
GNU General Public License v2.0
718 stars 208 forks source link

cloud not build cuda engine #289

Open Michaelzeyong opened 2 years ago

Michaelzeyong commented 2 years ago

I use win10 + cuda10.2 + 2080super, build tkdnn success! but when I try to transfer .rt files for yolov4 10241024, it saia "cloud not build cuda engine". yolov4 800800 success generate fp32 and fp16 .rt files, but only fp16 can use demo.exe. I cant solve this error. anyone else had fix it! thank you!

nuriakiin commented 1 year ago

I have same issue. Cant resolve it. Did you solve it? Screenshot from 2022-09-28 14-31-16

nuriakiin commented 1 year ago

I think is about memory or something same. I tried small models and worked. Then try again big model worked.