Add full operation to direct bindings

[rdspring1]

PR List

• Add threshold and clamp operations to direct_bindings #4876
• Add addcmul operation to direct_bindings #4877
• Add slice operation to direct bindings #4878
• Add Iota operation to direct bindings #4879
• Add full operation to direct bindings #4880 << This PR.
• Add topk operation to direct bindings #4881
• Support alias outputs in direct bindings. #4882
• Add welford operation to direct bindings #4883
• Add scatter operation to direct bindings #4884

[github-actions]

Review updated until commit 1059382

Description

• Added full operation to Python direct bindings

• Implemented shape handling for list/tuple inputs

• Added test case for scalar-only fusion scenarios

---

Changes walkthrough 📝

	Relevant files
Enhancement	ops.cpp Implement full op with shape flexibility                                  python/python_direct/ops.cpp Added template function full_op_fn for shape conversion Bound full operation with list/tuple shape support Added docstring for full operation +40/-0    python_translate.cpp Map FullOp to Python frontend                                                        python/python_direct/python_translate.cpp Added FullOp handler for Python frontend mapping Dispatched fill value creation for dynamic support Generated kwargs for shape, fill_value, and dtype +19/-0    opinfos.py Enable direct bindings for full op                                              tests/python/opinfo/opinfos.py Updated full_opinfo to support direct bindings Added supports_direct_bindings=True flag +1/-0
Tests	test_python_frontend.py Test full op with scalar inputs                                                    tests/python/direct/test_python_frontend.py Added test case test_scalar_only_inputs Verified full op with scalar addition in fusion Compared NVFuser output with PyTorch equivalent +20/-0

---

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🧪 PR contains tests

⚡ Recommended focus areas for review Shape Handling The SequenceAsVector function converts Python sequences to a vector of Val*. Ensure it properly handles edge cases like empty shapes, non-integer elements, or mixed types (e.g., tuple of lists). Testing for dynamic shapes (e.g., symbolic dimensions) is also critical. template <class ShapeType> TensorView* full_op_fn( ShapeType generic_output_shape, Val* fill_value, PrimDataType dtype) { std::vector<Val*> output_shape = SequenceAsVector(generic_output_shape); return full(output_shape, fill_value, dtype); } Fill Value Validity The dispatch(fop->getFillValue()) call may not validate dynamic fill values (e.g., symbolic scalars). Confirm that all valid input types for fill_value are supported and errors are handled gracefully. void handle(const FullOp* fop) final { NVF_ERROR(fop != nullptr); TensorView* out_tv = fop->output(0)->as<TensorView>(); visited_vals_.insert(out_tv); // Fill value can be dynamic so create it dispatch(fop->getFillValue()); static const std::vector<std::string> argument_names = { "shape", "fill_value", "dtype"}; printer_.generateKwargsOperation( "fd.ops.full", std::make_tuple(), argument_names, std::make_tuple(getShape(out_tv), fop->getFillValue(), out_tv->dtype()), {out_tv}); } Test Coverage The test test_scalar_only_inputs uses a fixed shape [2,2]. Consider adding tests for dynamic shapes, varying dtypes, and invalid inputs (e.g., negative dimensions) to ensure robustness. def test_scalar_only_inputs(nvfuser_direct_test): # We don't allow scalar outputs, currently, # so a tensor has to be returned def fusion_func(fd: FusionDefinition): s0 = fd.define_scalar() s1 = fd.define_scalar() s2 = fd.ops.add(s0, s1) c0 = fd.define_scalar(1.0, DataType.Float) t3 = fd.ops.full(shape=[2, 2], fill_value=c0, dtype=DataType.Float) t4 = fd.ops.mul(t3, s2) fd.add_output(t4) with FusionDefinition() as fd: fusion_func(fd) nvf_out, _ = nvfuser_direct_test.exec_nvfuser(fusion_func, [2.0, 3.0]) eager_out = torch.full([2, 2], 1.0) * 5.0 nvfuser_direct_test.assertEqual(eager_out, nvf_out[0])

[jjsjann123]

LGTM with minor nitpick on comment.

[rdspring1]

!build

[rdspring1]

!build