Pytorch_onnx_trt strange error

Description

converting model to trt error
nvinfer1::rt::cuda::WinogradConvActRunner::WinogradConvActRunner(nvinfer1::rt::DefaultRunnerParameters, const nvinfer1::ConvolutionParameters&, const std::vector&): Assertion `matchNbDims(mSrcTensors[0], mDstTensors[0]) && (mSrcTensors.size() == 1 || matchValidDims(popFront(mSrcTensors[1].extent), popFront(mDstTensors[0].extent)))’ failed.
Aborted (core dumped)

Environment

TensorRT Version: 7
GPU Type: 930M
Nvidia Driver Version: 440
CUDA Version: 10.2
CUDNN Version: 7.6
Operating System + Version: ubuntu18.04
Python Version (if applicable):
TensorFlow Version (if applicable):
PyTorch Version (if applicable):
Baremetal or Container (if container which image + tag):

Relevant Files

./trtexec --explicitBatch --onnx=model.onnx
&&&& RUNNING TensorRT.trtexec # ./trtexec --explicitBatch --onnx=model.onnx
[04/12/2020-15:53:38] [I] === Model Options ===
[04/12/2020-15:53:38] [I] Format: ONNX
[04/12/2020-15:53:38] [I] Model: model.onnx
[04/12/2020-15:53:38] [I] Output:
[04/12/2020-15:53:38] [I] === Build Options ===
[04/12/2020-15:53:38] [I] Max batch: explicit
[04/12/2020-15:53:38] [I] Workspace: 16 MB
[04/12/2020-15:53:38] [I] minTiming: 1
[04/12/2020-15:53:38] [I] avgTiming: 8
[04/12/2020-15:53:38] [I] Precision: FP32
[04/12/2020-15:53:38] [I] Calibration:
[04/12/2020-15:53:38] [I] Safe mode: Disabled
[04/12/2020-15:53:38] [I] Save engine:
[04/12/2020-15:53:38] [I] Load engine:
[04/12/2020-15:53:38] [I] Inputs format: fp32:CHW
[04/12/2020-15:53:38] [I] Outputs format: fp32:CHW
[04/12/2020-15:53:38] [I] Input build shapes: model
[04/12/2020-15:53:38] [I] === System Options ===
[04/12/2020-15:53:38] [I] Device: 0
[04/12/2020-15:53:38] [I] DLACore:
[04/12/2020-15:53:38] [I] Plugins:
[04/12/2020-15:53:38] [I] === Inference Options ===
[04/12/2020-15:53:38] [I] Batch: Explicit
[04/12/2020-15:53:38] [I] Iterations: 10
[04/12/2020-15:53:38] [I] Duration: 3s (+ 200ms warm up)
[04/12/2020-15:53:38] [I] Sleep time: 0ms
[04/12/2020-15:53:38] [I] Streams: 1
[04/12/2020-15:53:38] [I] ExposeDMA: Disabled
[04/12/2020-15:53:38] [I] Spin-wait: Disabled
[04/12/2020-15:53:38] [I] Multithreading: Disabled
[04/12/2020-15:53:38] [I] CUDA Graph: Disabled
[04/12/2020-15:53:38] [I] Skip inference: Disabled
[04/12/2020-15:53:38] [I] Inputs:
[04/12/2020-15:53:38] [I] === Reporting Options ===
[04/12/2020-15:53:38] [I] Verbose: Disabled
[04/12/2020-15:53:38] [I] Averages: 10 inferences
[04/12/2020-15:53:38] [I] Percentile: 99
[04/12/2020-15:53:38] [I] Dump output: Disabled
[04/12/2020-15:53:38] [I] Profile: Disabled
[04/12/2020-15:53:38] [I] Export timing to JSON file:
[04/12/2020-15:53:38] [I] Export output to JSON file:
[04/12/2020-15:53:38] [I] Export profile to JSON file:
[04/12/2020-15:53:38] [I]

Input filename: model.onnx
ONNX IR version: 0.0.4
Opset version: 9
Producer name: pytorch
Producer version: 1.3
Domain:
Model version: 0
Doc string:

[04/12/2020-15:53:39] [W] [TRT] onnx2trt_utils.cpp:198: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[04/12/2020-15:53:39] [W] [TRT] onnx2trt_utils.cpp:198: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[04/12/2020-15:53:39] [W] [TRT] onnx2trt_utils.cpp:198: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[04/12/2020-15:53:39] [W] [TRT] onnx2trt_utils.cpp:198: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[04/12/2020-15:53:39] [W] [TRT] onnx2trt_utils.cpp:198: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[04/12/2020-15:53:39] [W] [TRT] onnx2trt_utils.cpp:198: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[04/12/2020-15:53:39] [W] [TRT] Calling isShapeTensor before the entire network is constructed may result in an inaccurate result.
[04/12/2020-15:53:39] [W] [TRT] Calling isShapeTensor before the entire network is constructed may result in an inaccurate result.
[04/12/2020-15:53:39] [W] [TRT] Calling isShapeTensor before the entire network is constructed may result in an inaccurate result.
[04/12/2020-15:53:39] [W] [TRT] Unused Input: learned_1
[04/12/2020-15:53:39] [W] [TRT] Unused Input: learned_2
[04/12/2020-15:53:39] [W] [TRT] Unused Input: learned_3
[04/12/2020-15:53:46] [I] [TRT] Some tactics do not have sufficient workspace memory to run. Increasing workspace size may increase performance, please check verbose output.
trtexec: …/rtExt/cuda/customWinogradConvActRunner.cpp:48: nvinfer1::rt::cuda::WinogradConvActRunner::WinogradConvActRunner(nvinfer1::rt::DefaultRunnerParameters, const nvinfer1::ConvolutionParameters&, const std::vector&): Assertion `matchNbDims(mSrcTensors[0], mDstTensors[0]) && (mSrcTensors.size() == 1 || matchValidDims(popFront(mSrcTensors[1].extent), popFront(mDstTensors[0].extent)))’ failed.
Aborted (core dumped)

Steps To Reproduce

pytorch model converted to onnx after it using ./trtexec --explicitBatch --onnx=model.onnx to convert

Hi,

Could you please share the script and model file to reproduce this issue so we can better help?

Thanks

1 Like

how can i share model nvidia made my post hidden!

a

Thanks for sharing the model file, we will look into it and update you accordingly.

This kind of model is currently not supported.
For explicit precision mode, kernels weights must be build time.

Thanks

1 Like