Route Pytorch Conv1D/2D with groups == channels to depthwise conv

[HamzaEzzRa]

Description

Similar to the Keras issue solved in #1486, Pytorch Conv1D/2D where groups == in_channels == out_channels is not routed to the correct DepthwiseConv codegen path. Since Pytorch doesn't have a built-in DepthwiseConv layer, the only way I know to express this operation is to change a standard Conv groups to match in_channels and out_channels, which the pytorch_to_hls.py converter strictly rejects.

This patch relaxes the condition in pytorch_to_hls.py to include groups > 1 and groups == in_channels == out_channels as a valid configuration, then routes to the correct DepthwiseConv1D/2D graph layer inside convolution.py and reshapes to (depth_multiplier, n_chan, *kernel_size) with depth_multiplier = 1 for the depthwise conv.

Type of change

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

Tests

Added two tests:

• test/pytest/test_depthconv1d_pytorch.py
• test/pytest/test_depthconv2d_pytorch.py

Both are similar to the existing test_depthconv1d.py and test_depthconv2d.py, with code adapted from test_pytorch_api.py for Pytorch compatibility. Vivado and Vitis tests passed, but I don't have Catapult installed in my system to verify its backend.

Test Configuration:

Same pytest configuration as the existing test_depthconv1d.py and test_depthconv2d.py respectively.

pytest test/pytest/test_depthconv*d_pytorch.py

Checklist

• I have read the guidelines for contributing.
• I have commented my code, particularly in hard-to-understand areas.
• I have made corresponding changes to the documentation.
• My changes generate no new warnings.
• I have installed and run pre-commit on the files I edited or added.
• I have added tests that prove my fix is effective or that my feature works.