[torchlib] prims::sum

[justinchuby]

Fix pytorch/pytorch#173074

[Copilot]

Pull request overview

This PR implements the prims::sum operation to fix issue #173074. The implementation adds support for tensor sum reduction with optional dimension specification and output dtype casting.

Changes:

• Implements prims_sum function using ONNX ReduceSum operator with optional dtype casting
• Adds end-to-end test for torch.std_mean which decomposes into prims.sum

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
onnxscript/function_libs/torch_lib/ops/prims.py	Implements prims_sum function with ReduceSum and optional Cast operations
tests/function_libs/torch_lib/e2e_ops_tests.py	Adds test case for torch.std_mean to validate the prims.sum implementation

[Copilot]

The dims parameter should be normalized to None when it's an empty sequence, following the pattern used in prims_var (lines 839-841). ONNX's ReduceSum operator may not handle empty sequences correctly, and converting empty sequences to None ensures the operator reduces over all dimensions as expected.

Suggested change

	if dims is not None and len(dims) == 0:
	dims = None

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
12883	1	12882	1191

View the top 1 failed test(s) by shortest run time

onnxscript.tools.memory_peak_test.TestMemoryPeak::test_spy

Stack Traces | 0.057s run time

worker 'gw1' crashed while running 'onnxscript/tools/memory_peak_test.py::TestMemoryPeak::test_spy'

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}