PyTorch geometric quantization support

[i-riyad]

What does this PR do?

Type of change: New feature

Overview: Support quantization of PyTorch Geometric

# Add a code snippet demonstrating how to use this

Testing

python -m pytest tests/unit/torch/quantization/plugins/test_pytorch_geometric_plugin.py -v

Before your PR is "Ready for review"

• Make sure you read and follow Contributor guidelines and your commits are signed.
• Is this change backward compatible?: Yes
• Did you write any new necessary tests?: Yes
• Did you add or update any necessary documentation?: Yes
• Did you update Changelog?: Yes

Additional Information

• Support for quantize the graph neural networks (GNNs) #298

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 73.43%. Comparing base (72f23dc) to head (5386681).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #494   +/-   ##
=======================================
  Coverage   73.43%   73.43%           
=======================================
  Files         180      180           
  Lines       18146    18146           
=======================================
  Hits        13326    13326           
  Misses       4820     4820           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[kevalmorabia97]

Exact date might get pushed

Suggested change

	0.40 (2025-12-09)
	0.40 (2025-12-xx)

[kevalmorabia97]

LGTM on a high level. Approving as codeowner

[realAsma]

nit: This might not be needed

Suggested change

	def forward(self, input, *args, **kwargs):
	"""Forward pass with quantization.
	Args:
	input: Input tensor to the linear layer
	*args: Additional positional arguments
	**kwargs: Additional keyword arguments
	Returns:
	Quantized output tensor
	"""
	# Quantize input activations
	input_q = self.input_quantizer(input)
	# Quantize weights
	weight_q = self.weight_quantizer(self.weight)
	# Perform linear operation
	output = torch.nn.functional.linear(
	input_q,
	weight_q,
	self.bias if hasattr(self, "bias") and self.bias is not None else None,
	)
	# Quantize output (typically disabled by default)
	return self.output_quantizer(output)

[realAsma]

For example, see the definition for ConvNd layers:

TensorRT-Model-Optimizer/modelopt/torch/quantization/nn/modules/quant_conv.py

Line 39 in ed58324

@QuantModuleRegistry.register({nn.Conv1d: "nn.Conv1d"})

The forward path should work (inherited from QuantLinearConvBase)

[realAsma]

nit:

For debugging, I directly run the unitests using VSCode/Cursor test explorer.

Suggested change

	if __name__ == "__main__":
	pytest.main([__file__])