CAST operator in ONNX model to convert INT32 to other types for DLA

Autonomous Machines Jetson & Embedded Systems Jetson AGX Orin
, , ,
1

Description

im converting pytorch to onnx model then to tensorrt. while converting to trt, i got these errors:
-Please explicitly use CAST operator in ONNX model or add an identity layer to convert INT32 to other types for DLA.
-DLA Layer /model.0/conv/Conv does not support dynamic shapes in any dimension.
-DLA only supports FP16 and Int8 precision type. Switching this layer’s device type to GPU.

i added code lines for casting to fp16 on the pt2onnx file, but i still get the errors

Environment

TensorRT Version: 8.2.5.1
GPU Type: Jetson AGX Orin
Nvidia Driver Version: 35.1.0
CUDA Version: 11.4
CUDNN Version: 8.9.4.25
Operating System + Version: Ubuntu 20.04
PyTorch Version (if applicable): 1.14

Relevant Files

pt to onnx code:
pt2onnx.py (5.7 KB)
onnx to trt trtexec file:
onnx2trt.sh (704 Bytes)

Thank you

4

Hi,
Please check the below links, as they might answer your concerns.

We are moving this post to the Jetson AGX Orin forum to get better help.

Thanks!

5

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