12 // Returns "Tensor['N', 'C', 'H', 'W']" for a tensor with names ('N', 'C', 'H', 'W').
15   os << "Tensor" << tensor.names();  in toDimnameRepr()
25   const auto names = tensor.names();  in dimname_to_position()  local
27   const auto it = std::find(names.begin(), names.end(), dim);  in dimname_to_position()
28   TORCH_CHECK(it != names.end(),  in dimname_to_position()
31   return std::distance(names.begin(), it);  in dimname_to_position()
46     DimnameList names,  in report_positional_error()  argument
49 …// TODO(zou3519): Can improve message by checking if names are alignable and suggesting workarounds  in report_positional_error()
51       "Error when attempting to ", action, " dims ", names, " and dims ",  in report_positional_error()
58     DimnameList names,  in check_for_misalignment()  argument
65 …// TODO(zou3519): Can improve message by checking if names are alignable and suggesting workarounds  in check_for_misalignment()
67       "Misaligned dims when attempting to ", action, " dims ", names, " and dims ",  in check_for_misalignment()
72 // Assumption: A DimnameList can have no duplicate full names with
75     DimnameList names,  in unify_from_right()  argument
79   const auto size = std::max(names.size(), other_names.size());  in unify_from_right()
82   auto names_it = names.rbegin();  in unify_from_right()
85   while (names_it != names.rend() || other_it != other_names.rend()) {  in unify_from_right()
86     const auto& name = names_it == names.rend() ? wildcard : *names_it;  in unify_from_right()
89     // Step 1: Check that the names match  in unify_from_right()
92       report_positional_error(name, other_name, names, other_names, action);  in unify_from_right()
96     // Step 2: Check that the names are not misaligned  in unify_from_right()
98       // Let: N = max(len(names), len(other_names))  in unify_from_right()
99       //      K = # of special names among names and other_names.  in unify_from_right()
101       check_for_misalignment(name, names, other_names, action);  in unify_from_right()
102       check_for_misalignment(other_name, other_names, names, action);  in unify_from_right()
105     if (names_it != names.rend()) {  in unify_from_right()
127       "Name mismatch: specified out tensor with names ", a,  in assert_names_equal()
128       " are not the same as the computed output names ", b,  in assert_names_equal()
156 const Tensor& propagate_names(const Tensor& result, DimnameList names, bool validate_names) {  in propagate_names()  argument
157   propagate_names(result.unsafeGetTensorImpl(), names, validate_names);  in propagate_names()
161 TensorImpl* propagate_names(TensorImpl* result, DimnameList names, bool validate_names) {  in propagate_names()  argument
164         !names.empty(),  in propagate_names()
165         "propagate_names: passed in empty names to propagate to result with",  in propagate_names()
166         " shape ", result->sizes(), ". Empty names means that name inference did",  in propagate_names()
170     impl::internal_set_names_inplace(result, names, validate_names);  in propagate_names()
172     assert_names_equal(impl::get_names(result), names);  in propagate_names()
181   const auto src_names = src.names();  in propagate_names_except()
237   auto tensor_names = tensor.names();  in compute_squeeze_outnames()
251   auto tensor_names = tensor.names();  in compute_squeeze_outnames()
268   auto tensor_names = tensor.names();  in compute_diagonal_outnames()
293     " would produce output tensor with duplicate names ",  in check_feature_names_are_distinct()
298 static int64_t num_batch_dims(DimnameList names) {  in num_batch_dims()  argument
299   if (names.size() <= 2) {  in num_batch_dims()
302   return static_cast<int64_t>(names.size() - 2);  in num_batch_dims()
318   // To compute output names, we unify the batch dimensions because those are  in compute_matmul_outnames()
321   // After that, we append some names that are equal to the result of the matmul  in compute_matmul_outnames()
322   // without batch dimensions. Those names are computed by removing the names  in compute_matmul_outnames()
326   // Get the output's batch dimension names  in compute_matmul_outnames()
333   // completely contracted away during matmul so we don't take any names from them.  in compute_matmul_outnames()
354   auto mv_outnames = compute_matmul_outnames(mat.names(), vec.names());  in propagate_names_for_addmv()
355   return unify_from_right(mv_outnames, bias.names());  in propagate_names_for_addmv()
367   auto mm_outnames = compute_matmul_outnames(m1.names(), m2.names());  in propagate_names_for_addmm()
368   return unify_from_right(mm_outnames, bias.names());  in propagate_names_for_addmm()
407   return unify_from_right(self.names(), other.names());  in compute_broadcast_outnames()
417   auto reference_names = reference_tensor.names();  in broadcast_to_outnames()
418   auto tensor_names = tensor.names();  in broadcast_to_outnames()
434     const auto tensor_names = tensor.names();  in compute_cat_outnames()
450   return compute_matmul_outnames(self.names(), other.names());  in compute_matmul_outnames()
459   const auto self_names = self.names();  in compute_cdist_outnames()
460   const auto other_names = other.names();  in compute_cdist_outnames()
470   // distance values. We propagate the names of the dimension of size M (in self)  in compute_cdist_outnames()
486   return compute_matmul_outnames(self.names(), other.names());  in compute_bmm_outnames()
498   auto bmm_names = compute_matmul_outnames(self.names(), other.names());  in compute_baddbmm_outnames()
499   auto baddbmm_names = unify_from_right(bias.names(), bmm_names);  in compute_baddbmm_outnames()