Cannot convert onnx to engine

[AIisCool]

[12/15/2023-19:46:53] [V] [TRT] Registering layer: Add_121 for ONNX node: Add_121
[12/15/2023-19:46:53] [V] [TRT] Registering tensor: 186 for ONNX tensor: 186
[12/15/2023-19:46:53] [V] [TRT] Add_121 [Add] outputs: [186 -> (-1, 64, -1, -1)[FLOAT]], 
[12/15/2023-19:46:53] [V] [TRT] Static check for parsing node: Add_122 [Add]
[12/15/2023-19:46:53] [V] [TRT] Parsing node: Add_122 [Add]
[12/15/2023-19:46:53] [V] [TRT] Searching for input: 186
[12/15/2023-19:46:53] [V] [TRT] Searching for input: 65
[12/15/2023-19:46:53] [V] [TRT] Add_122 [Add] inputs: [186 -> (-1, 64, -1, -1)[FLOAT]], [65 -> (-1, 64, -1, -1)[FLOAT]], 
[12/15/2023-19:46:53] [V] [TRT] Registering layer: Add_122 for ONNX node: Add_122
[12/15/2023-19:46:53] [V] [TRT] Registering tensor: 187 for ONNX tensor: 187
[12/15/2023-19:46:53] [V] [TRT] Add_122 [Add] outputs: [187 -> (-1, 64, -1, -1)[FLOAT]], 
[12/15/2023-19:46:53] [V] [TRT] Static check for parsing node: Conv_123 [Conv]
[12/15/2023-19:46:53] [V] [TRT] Parsing node: Conv_123 [Conv]
[12/15/2023-19:46:53] [V] [TRT] Searching for input: 187
[12/15/2023-19:46:53] [V] [TRT] Searching for input: m_tail.weight
[12/15/2023-19:46:53] [V] [TRT] Conv_123 [Conv] inputs: [187 -> (-1, 64, -1, -1)[FLOAT]], [m_tail.weight -> (3, 64, 3, 3)[FLOAT]], 
[12/15/2023-19:46:53] [V] [TRT] Convolution input dimensions: (-1, 64, -1, -1)
[12/15/2023-19:46:53] [V] [TRT] Registering layer: Conv_123 for ONNX node: Conv_123
[12/15/2023-19:46:53] [V] [TRT] Using kernel: (3, 3), strides: (1, 1), prepadding: (1, 1), postpadding: (1, 1), dilations: (1, 1), numOutputs: 3
[12/15/2023-19:46:53] [V] [TRT] Convolution output dimensions: (-1, 3, -1, -1)
[12/15/2023-19:46:53] [V] [TRT] Registering tensor: output_0 for ONNX tensor: output
[12/15/2023-19:46:53] [V] [TRT] Conv_123 [Conv] outputs: [output -> (-1, 3, -1, -1)[FLOAT]], 
[12/15/2023-19:46:53] [V] [TRT] Marking output_0 as output: output
[12/15/2023-19:46:53] [I] Finished parsing network model. Parse time: 2.17735
[12/15/2023-19:46:53] [I] Set shape of input tensor input for optimization profile 0 to: MIN=1x3x574x706 OPT=1x3x574x706 MAX=1x3x574x706
[12/15/2023-19:46:53] [V] [TRT] Trying to set exclusive file lock C:/VapourSynth/vsmlrt/models/dpir\f82cfa0a.engine.cache.lock

[12/15/2023-19:46:53] [W] [TRT] Could not read timing cache from: C:/VapourSynth/vsmlrt/models/dpir\f82cfa0a.engine.cache. A new timing cache will be generated and written.
[12/15/2023-19:46:53] [E] Error[4]: [network.cpp::nvinfer1::Network::validate::3640] Error Code 4: Internal Error (input: for dimension number 1 in profile 0 does not match network definition (got min=3, opt=3, max=3), expected min=opt=max=4).)
[12/15/2023-19:46:53] [E] Engine could not be created from network
[12/15/2023-19:46:53] [E] Building engine failed
[12/15/2023-19:46:53] [E] Failed to create engine from model or file.
[12/15/2023-19:46:53] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec [TensorRT v9100] # C:/VapourSynth/vsmlrt\vsmlrt-cuda\trtexec --onnx=C:/VapourSynth/vsmlrt/models/dpir/drunet_deblocking_color.onnx

[WolframRhodium]

DPIR is different in that it requires an additional input to specify the noise strength, so the number of input channel is 4 instead of 3.

In addition, it also requires dimensions to be multiples of 8, so it should be trtexec --shapes=1x4x576x712.