Trtexec --onnx --saveEngine --fp16 error

Autonomous Machines Jetson & Embedded Systems Jetson AGX Orin
1 
/usr/src/tensorrt/bin/trtexec --onnx=yolov5m.onnx --shapes=images:1x3x640x640 --fp16    --saveEngine=engine

[09/17/2022-23:41:05] [W] [TRT] Weights [name=Conv_30 + PWN(PWN(Sigmoid_31), Mul_32).weight] had the following issues when converted to FP16:
[09/17/2022-23:41:05] [W] [TRT]  - Subnormal FP16 values detected. 
[09/17/2022-23:41:05] [W] [TRT] If this is not the desired behavior, please modify the weights or retrain with regularization to reduce the magnitude of the weights.
[09/17/2022-23:41:05] [W] [TRT] Weights [name=Conv_30 + PWN(PWN(Sigmoid_31), Mul_32).weight] had the following issues when converted to FP16:
[09/17/2022-23:41:05] [W] [TRT]  - Subnormal FP16 values detected. 
[09/17/2022-23:41:05] [W] [TRT] If this is not the desired behavior, please modify the weights or retrain with regularization to reduce the magnitude of the weights.
[09/17/2022-23:41:05] [W] [TRT] Weights [name=Conv_30 + PWN(PWN(Sigmoid_31), Mul_32).weight] had the following issues when converted to FP16:
[09/17/2022-23:41:05] [W] [TRT]  - Subnormal FP16 values detected. 
[09/17/2022-23:41:05] [W] [TRT] If this is not the desired behavior, please modify the weights or retrain with regularization to reduce the magnitude of the weights.
[09/17/2022-23:41:06] [W] [TRT] Weights [name=Conv_30 + PWN(PWN(Sigmoid_31), Mul_32).weight] had the following issues when converted to FP16:
[09/17/2022-23:41:06] [W] [TRT]  - Subnormal FP16 values detected. 
[09/17/2022-23:41:06] [W] [TRT] If this is not the desired behavior, please modify the weights or retrain with regularization to reduce the magnitude of the weights.
3

Hi,

These are just some warning messages to let you know the TensorRT handling in case it doesn’t meet your expectation.

Thanks.

5

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