Warning While DS App Converts ONNX Model To Engine

Accelerated Computing Intelligent Video Analytics DeepStream SDK
#1

dGPU
NVIDIA Driver：510.73.05
CUDA Toolkit：CUDA 11.6 Update 1
cuDNN：8.4.0.27
TensorRT：8.2.5.1
Deepstream 6.1

Sample of the warning prints-

WARNING: [TRT]: TensorRT was linked against cuDNN 8.4.1 but loaded cuDNN 8.4.0
WARNING: [TRT]: Weights [name=Conv_0 + Relu_1.weight] had the following issues when converted to FP16:
WARNING: [TRT]:  - Subnormal FP16 values detected. 
WARNING: [TRT]: If this is not the desired behavior, please modify the weights or retrain with regularization to reduce the magnitude of the weights.
WARNING: [TRT]: Weights [name=Conv_0 + Relu_1.weight] had the following issues when converted to FP16:
WARNING: [TRT]:  - Subnormal FP16 values detected. 
WARNING: [TRT]: If this is not the desired behavior, please modify the weights or retrain with regularization to reduce the magnitude of the weights.
WARNING: [TRT]: Weights [name=Conv_0 + Relu_1.weight] had the following issues when converted to FP16:
WARNING: [TRT]:  - Subnormal FP16 values detected.

I have attached
DS6_1_ONNX_TRT_Warns.txt (592.3 KB)
a txt file with the all the warnings during conversion.

#3

It seems it exceeds FP16 MIN MAX value.

#4

So for this particular ONNX file, we should run it in FP32 mode only?

#5

Yes, or it will have precision drop.

#7

There is no update from you for a period, assuming this is not an issue anymore.
Hence we are closing this topic. If need further support, please open a new one.
Thanks

#8

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.