Onnx-tensorrt: Assertion failed: ctx->tensors().count(inputName)

[huangrichen11]

When I convert onnx file to trt engine, but got this error below:
while parsing node number 224 [Resize]:
Assertion failed: cxt->tensors().count(inputName)
When I use onnxsim to simplify onnx to merge constant node, but i got this error below:
In node-1 (parseGraph): unsupported _node: Assertion failed: convertOnnxWeights(initialize, &weights, ctx)

version:

• trt 7.0
• onnx 1.6.0
• pytorch 1.4

[Syencil]

I have the same problem.

While parsing node number 362 [Resize]:
ERROR: ModelImporter.cpp:124 In function parseGraph:
[5] Assertion failed: ctx->tensors().count(inputName)

When i visiualized onnx model by netron, it seem have a constant op before resize op

[nigyiii]

Did you solve this problem? I met the same error.
trt 7.0
pytorch 1.4.0
interpolate mode: bilinear

[Syencil]

i have solved it.
just recompile the newest onnx lib and set opset=11 in trt7. hope it works

[diandianliu]

i have solved it.
just recompile the newest onnx lib and set opset=11 in trt7. hope it works

I have the same problem, what's the onnx lib name that your replace? thanks

[wdhxek]

Hi, @Syencil .
same problem faced when I tried to convert a onnx model into trt engine using trtexec.exe.
Could you tell me more details for solving this problem, as I can not find any onnx lib. Do you mean nvonnxparser.dll or nvparsers.dll (windows pakage download from nvidia website) ?

Any suggestion is appreciated, tks!

Input filename: 1.onnx
ONNX IR version: 0.0.6
Opset version: 11
Producer name: pytorch
Producer version: 1.5
cuda 10.2
cudnn 7.6

tensorrt 7.0.1

While parsing node number 169 [Resize]:
ERROR: ModelImporter.cpp:124 In function parseGraph:
[5] Assertion failed: ctx->tensors().count(inputName)
[07/30/2020-15:39:52] [E] Failed to parse onnx file

[xizi]

I have the same problem.
pytorch 1.4.0 or 1.5.0
onnx 1.6.0 or 1.7.0
tensorrt 7.0.0

[baowenbo]

i have solved it.
just recompile the newest onnx lib and set opset=11 in trt7. hope it works

We are appreciated that you may provide some detailed help on this.

[wanghr323]

i have solved it.
just recompile the newest onnx lib and set opset=11 in trt7. hope it works

We are appreciated that you may provide some detailed help on this.

I have solved the problem,
first,
replace your parser/onnx dir with 'https://github.com/onnx/onnx-tensorrt/tree/7.0 '(the newset brach)
and replace the parser/onnx/third_party/onnx with 'https://github.com/onnx/onnx/tree/master' (master branch)

second,
edit one file following:
You could try editing the ONNX parser source like this PR, building it, and seeing if it parses your model successfully: https://github.com/onnx/onnx-tensorrt/pull/418/files
(ref:https://github.com/NVIDIA/TensorRT/issues/422)

rebuild your tensorRT project and replace the libnvonnxparser.so of your system.
it works.

[kevinch-nv]

This issue should be solved in the later TensorRT releases. Please upgrade and see if the new releases solves the issue.

If not, feel free to open a new one.

[shirleyxy]

This issue should be solved in the later TensorRT releases. Please upgrade and see if the new releases solves the issue.

If not, feel free to open a new one.

I have faced this problem while parsing my .onnx model using tensorrt7.2.2.3 trtexec.exe.
onnx 1.6.0 pytorch 1.3.1

[kevinch-nv]

@shirleyxy can you open an issue with the model you are having trouble importing?

[whcjb]

the same problem, pytorch1.6, onnx1.8, tensorrt7.0