Why convolution can't offload to DLA?

When I convert my onnx model to TensorRT engine plan using trtexec with DLACore 0, but some convloution run on GPU which is supported by DLA.
What’s wrong with this?


Could you share the complete log with us?
Usually, the log will show why the layer fallbacks to GPU.


I have saw the log, there is no exception about the Conv_107.
Hope you can find some useful infomation.
convert_dla.log (3.9 MB)

Dear @distant1219,
I see below message in log “{ForeignNode[Conv_107…Reshape_239 + Transpose_240]} cannot be compiled by DLA, falling back to GPU.”

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