onnx->trt fail on mobilenet

VertexC / dl-infer-perf

deep learning inference perf analysis

2 stars 1 forks source link

onnx->trt fail on mobilenet #3

Open VertexC opened 3 years ago

VertexC commented 3 years ago

root@ubuntu1804-lts-base:/scratch/dev/dev/dl-infer-perf# python3 infer_perf/onnx2trt.py mobilenet
[TensorRT] ERROR: Network must have at least one output
Traceback (most recent call last):
  File "infer_perf/onnx2trt.py", line 80, in <module>
    runner = onnx2trt_runner(args.model, batch_size=args.batch)
  File "infer_perf/onnx2trt.py", line 60, in onnx2trt_runner
    context = engine.create_execution_context()
AttributeError: 'NoneType' object has no attribute 'create_execution_context'

VertexC commented 3 years ago

The error has something to do onnx's model version and current trt verison. Need further investigation.

VertexC commented 3 years ago

Seems to be a known issue. Backlog for now. https://forums.developer.nvidia.com/t/whether-tensorrt-supports-mobilenet-in-onnx-framework/80397