GPU not on full used. #156

Open
opened 2021-11-20 16:33:08 +01:00 by whenupdate · 2 comments
whenupdate commented 2021-11-20 16:33:08 +01:00 (Migrated from github.com)

I have try two different GPU, low GPU (6GB) and medium GPU (16GB) but always stuck on 2 iterations. I check GPU monitoring resource low and medium GPU have same perform, yeah its stuck on 2GB GPU used. When on last update before this stuck i always got 4-6 iteration on that GPU. This is my version.

python 3.8
Cuda 11
onnx last update
onnxruntime-gpu last update
insightface==0.2.1

  • Are someone trying to solve this issue ?
  • Downgrade python and cuda is not solve the issue ?
  • Or maybe this repo need update on supp onnx or cuda ?
I have try two different GPU, low GPU (6GB) and medium GPU (16GB) but always stuck on 2 iterations. I check GPU monitoring resource low and medium GPU have same perform, yeah its stuck on 2GB GPU used. When on last update before this stuck i always got 4-6 iteration on that GPU. This is my version. python 3.8 Cuda 11 onnx last update onnxruntime-gpu last update insightface==0.2.1 - Are someone trying to solve this issue ? - Downgrade python and cuda is not solve the issue ? - Or maybe this repo need update on supp onnx or cuda ?
nonlin commented 2021-12-10 06:50:01 +01:00 (Migrated from github.com)

I noticed the same thing. I'm using TensorRT and CUDA with ONNX.
My GPU Load hits 17% and VRAM uses 4GB.

Sometimes my it/s will go from 4 to 16 IDK what makes it spike.

I noticed the same thing. I'm using TensorRT and CUDA with ONNX. My GPU Load hits 17% and VRAM uses 4GB. Sometimes my it/s will go from 4 to 16 IDK what makes it spike.
johnrom1243 commented 2022-12-08 05:33:05 +01:00 (Migrated from github.com)

did you ever manage to find a solution?

did you ever manage to find a solution?
Sign in to join this conversation.