addmm: error on output dtype mismatch. (#138520)

Pull Request resolved: https://github.com/pytorch/pytorch/pull/138520
Approved by: https://github.com/ezyang
ghstack dependencies: #138515

test/test_ops.py

@@ -123,8 +123,6 @@ aten = torch.ops.aten
 meta_consistency_out_dtype_mismatch_xfails = {
     xfail("abs"),
     xfail("addbmm"),
-    xfail("addmm"),
-    xfail("addmm", "decomposed"),
     xfail("addmv"),
     xfail("alias_copy"),
     xfail("all"),

torch/_decomp/decompositions.py

@@ -1479,7 +1479,7 @@ def tensor_split_tensor_indices_or_sections_py_impl(
 
 # TODO: this doesn't appear to have enough precision in bfloat16
 @register_decomposition(aten.addmm)
-@out_wrapper()
+@out_wrapper(exact_dtype=True)
 @pw_cast_for_opmath
 def addmm(self: Tensor, mat1: Tensor, mat2: Tensor, beta: int = 1, alpha: int = 1):
     if not self.is_floating_point() and not self.is_complex():