ONNX export mode is set to TrainingMode.EVAL, but operator 'instance_norm' is set to train=True
edwardnguyen1705 opened this issue · comments
Thanh Nguyen commented
Dear @jaybdub ,
I have tried torch2trt to convert an OSNet-AIN model based on fast-reid, and I observe the following WARNING
UserWarning: ONNX export mode is set to TrainingMode.EVAL, but operator 'instance_norm' is set to train=True. Exporting with train=True.
It seems that the warning could lead to the difference between outputs of pytorch and tensorrt are quite different on a person reid dataset. I attach the scripts to produce the results.
torch2trt.zip
Have encountered this issue and figured out a way to tackle this?
Thanks for your time.
jayong-sv commented
@edwardnguyen1705 I encountered the same error. Did you solve it?
Thanh Nguyen commented
Hi @jayong-sv ,
I made a mistake, actually the warning does not affect the result.