[WIP] Sparse addmm on SparseXPU

[jenniew]

Sparse ops on SparseXPU:
sparse addmm

Related issue: #2211

[CuiYifeng]

Main part of this PR LGTM.

[CuiYifeng]

Will Tensor mat1_dense = sparse_._to_dense() cause undefined issues? I noticed that dtype and masked are optional for sparse_to_dense.

[CuiYifeng]

xpu::check_device can be found in sparse/xpu/sycl/SparseTensorMathKernels.cpp.

[Copilot]

Pull request overview

This PR implements sparse addmm operations for XPU devices by adding support for addmm operations on sparse tensors. The implementation follows the existing PyTorch sparse tensor patterns by converting sparse tensors to dense format and delegating to the standard addmm operations.

Key changes:

• Added three addmm function variants (standard, out-variant, and in-place) to the native functions YAML with SparseXPU dispatch keys
• Implemented the sparse addmm operations in C++ by converting sparse tensors to dense format
• Removed the test skip decorator to enable existing sparse addmm tests for XPU

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
yaml/native/native_functions.yaml	Registered three addmm function variants with SparseXPU dispatch mappings
test/xpu/test_sparse_xpu.py	Removed skipXPUIf decorator to enable sparse addmm tests
src/ATen/native/sparse/xpu/SparseTensorMath.cpp	Implemented sparse addmm functions with device validation and sparse-to-dense conversion

Comments suppressed due to low confidence (1)

yaml/native/native_functions.yaml:1

• Corrected spelling of 'sprase' to 'sparse' in the skip message.

- func: copy(Tensor self, Tensor src, bool non_blocking=False) -> Tensor

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

This commented-out device check should either be implemented or removed. If device validation is needed beyond the individual tensor checks above, uncomment and ensure the function exists; otherwise, remove the dead code.

Suggested change

// TORCH_CHECK(xpu::check_device({sparse_, r_, t, dense}));

[Copilot]

Converting the sparse tensor to dense format defeats the purpose of sparse operations and may cause significant memory overhead for large sparse tensors. Consider implementing a proper sparse matrix multiplication kernel instead of this fallback approach.

Suggested change

	Tensor mat1_dense = sparse_._to_dense(std::nullopt, std::nullopt);
	at::addmm_out(r_, t, mat1_dense, dense, beta, alpha);
	// Use a proper sparse matrix multiplication kernel for XPU
	xpu::addmm_out_sparse_dense_kernel(r_, t, sparse_, dense, beta, alpha);