fix: conv parameter check failure

[chohk88]

Description

This PR fixes an issue where padding values given as a single-element list (e.g., [0] and [1]) were not properly converted in convNd, leading to a TensorRT error:

ERROR:torch_tensorrt [TensorRT Conversion Context]:IConvolutionLayer::setPaddingNd: Error Code 3: API Usage Error (Parameter check failed, condition: (padding.nbDims == 2 || padding.nbDims == 3) && allDimsGtEq(padding, 0) && allDimsLtEq(padding, kMAX_PADDING). )

For the padding value [1], this also caused shape mismatches, such as:

AssertionError: The values for attribute 'shape' do not match: torch.Size([1, 6, 32, 32]) != torch.Size([1, 6, 34, 34]).

If padding is [0], it still triggers a TensorRT paramater check failure error, but the final output is correct. This is a minor issue that doesn’t affect functionality.

Fix

• Now, if padding is a list with one element, it is expanded properly ((1,1) for conv2d, (1,1,1) for conv3d, etc.).
• Verified with additional test cases

Fixes #(issue)

Type of change

Please delete options that are not relevant and/or add your own.

• Bug fix (non-breaking change which fixes an issue)

Checklist:

• My code follows the style guidelines of this project (You can use the linters)
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas and hacks
• I have made corresponding changes to the documentation
• I have added tests to verify my fix or my feature
• New and existing unit tests pass locally with my changes
• I have added the relevant labels to my PR in so that relevant reviewers are notified

[zewenli98]

Good catch! Can you confirm if stride or dilation need to do the same changes or not?

[chohk88]

Good catch! Can you confirm if stride or dilation need to do the same changes or not?

Good catch! I missed that.

The same error occurs for list_dilation, e.g., param("list_dilation", 2, dilation=[3]), just like with padding. I have updated the code to properly handle list inputs for dilation and stride, ensuring they are correctly converted into tuples for the conv layer.

Additionally, I have revised the method for checking 1D convolution.

[zewenli98]

cool, LGTM!