Seek Help!!!onnx 3D Conv layer trans issue ../builder/cudnnBuilderUtils.cpp (354)

I modify some codes in onnx-tensorrt project to apply for my onnx net. when
run
auto trt_engine = common::infer_object(trt_builder->buildCudaEngine(*trt_network.get()));

some errors appear,what can i do for this problem

[2019-09-24 07:45:31 WARNING] Setting layouts of network and plugin input/output tensors to linear, as 3D operators are found and 3D non-linear IO formats are not supported, yet.
[2019-09-24 07:45:35 ERROR] …/builder/cudnnBuilderUtils.cpp (354) - Cuda Error in findFastestTactic: 700 (an illegal memory access was encountered)
[2019-09-24 07:45:35 ERROR] …/rtSafe/safeRuntime.cpp (32) - Cuda Error in free: 700 (an illegal memory access was encountered)

Hello. Same for me here when using this bit of code.

import tensorrt as trt
import onnx
import numpy as np

ONNX_MODEL = "torch_model.onnx"


def build_engine():
    with trt.Builder(
        TRT_LOGGER
    ) as builder, builder.create_network() as network, trt.OnnxParser(
        network, TRT_LOGGER
    ) as parser:
        # Configure the builder here.
        builder.max_workspace_size = 2 ** 30
        # Parse the model to create a network.
        with open(ONNX_MODEL, "rb") as model:
            parser.parse(model.read())
        # Build and return the engine. Note that the builder, network and parser are destroyed when this function returns.
        return builder.build_cuda_engine(network)

build_engine()

[TensorRT] WARNING: Setting layouts of network and plugin input/output tensors to linear, as 3D operators are found and 3D non-linear IO formats are not supported, yet.

I’m working with TensorRT 6.0.1.5, I thought it was the whole point, that 3D operators are supported now ?

I get the same error by using plain TensorRT with a 3D convolution with kernel size = 3 and strides = 2 and padding Same UPPER

Same problem as me.

same problem as me, is there any one solves it !

Hi,

Please refer below dev talk issue:
https://devtalk.nvidia.com/default/topic/1064564/tensorrt/dynamic-input-and-convolution

Thanks