| /external/pytorch/torch/fx/experimental/ |
| D | partitioner_utils.py | 31 input_nodes: Dict[Node, None] = {}
32 map_arg(node.args, input_nodes.setdefault)
33 map_arg(node.kwargs, input_nodes.setdefault)
35 for n in input_nodes:
46 input_nodes: Dict[Node, None] = {}
47 map_arg(node.args, input_nodes.setdefault)
48 map_arg(node.kwargs, input_nodes.setdefault)
52 for input_node in input_nodes:
107 input_nodes: Dict[Node, None] = {}
108 map_arg(node.args, input_nodes.setdefault)
[all …]
|
| /external/pytorch/torch/_inductor/codegen/cuda/ |
| D | gemm_template.py | 385 input_nodes: List[Buffer],
400 super().__init__("cutlass_gemm", input_nodes, layout, input_reorder)
403 assert len(input_nodes) == 2 or len(input_nodes) == 3
405 [node.get_layout() for node in input_nodes]
413 input_nodes: List[Buffer],
485 input_nodes: List[Buffer],
514 input_layouts = [node.get_layout() for node in input_nodes]
515 input_strides = [node.get_stride() for node in input_nodes]
744 X = self.input_nodes[0]
745 W = self.input_nodes[1]
[all …]
|
| D | cuda_template.py | 29 input_nodes: List[Buffer],
45 self.input_nodes = input_nodes
82 else list(range(len(self.input_nodes)))
85 unique(self.input_nodes[idx].get_name() for idx in input_reorder)
101 input_tensor_meta=TensorMeta.from_irnodes(self.input_nodes),
126 self.input_nodes,
|
| /external/pytorch/torch/_inductor/codegen/rocm/ |
| D | ck_universal_gemm_template.py | 84 input_nodes: List[Buffer],
92 input_nodes=input_nodes,
135 metas = [T.get_layout() for T in [*self.input_nodes, self.output_node]]
253 X, W = self.input_nodes[0], self.input_nodes[1]
255 Bias = self.input_nodes[2] if 3 == len(self.input_nodes) else None
330 T.get_layout() for T in [*self.input_nodes, self.output_node]
386 input_nodes, argument
395 input_nodes,
409 X = self.input_nodes[0]
410 W = self.input_nodes[1]
[all …]
|
| D | rocm_template.py | 27 input_nodes: List[Buffer],
43 self.input_nodes = input_nodes
81 else list(range(len(self.input_nodes)))
84 unique(self.input_nodes[idx].get_name() for idx in input_reorder)
100 input_tensor_meta=TensorMeta.from_irnodes(self.input_nodes),
125 self.input_nodes,
|
| /external/libaom/av1/encoder/ |
| D | ml.c | 31 void av1_nn_predict_c(const float *input_nodes, in av1_nn_predict_c() argument
50 val += layer_weights[node * num_input_nodes + i] * input_nodes[i]; in av1_nn_predict_c()
56 input_nodes = output_nodes; in av1_nn_predict_c()
66 val += layer_weights[node * num_input_nodes + i] * input_nodes[i]; in av1_nn_predict_c()
123 const float *input_nodes = feature; in av1_nn_predict_v2() local
129 input_nodes = nn_fc_forward(input_nodes, nn_config->layer + i); in av1_nn_predict_v2()
135 input_nodes = nn_fc_forward(input_nodes, nn_config->layer + num_layers); in av1_nn_predict_v2()
138 memcpy(output, input_nodes, sizeof(*input_nodes) * nn_config->num_logits); in av1_nn_predict_v2()
|
| /external/pytorch/torch/_inductor/ |
| D | select_algorithm.py | 122 input_nodes, argument
143 self.input_nodes = input_nodes
204 for i, inp in enumerate(itertools.chain(self.input_nodes, (self.output_node,))):
267 named_args = self.input_nodes[
268 self.prefix_args : len(self.input_nodes) - self.suffix_args
275 len(self.input_nodes),
278 for input_node in self.input_nodes[: self.prefix_args]:
297 for input_node in self.input_nodes[len(self.input_nodes) - self.suffix_args :]:
459 self.input_nodes[: self.prefix_args],
460 self.input_nodes[len(self.input_nodes) - self.suffix_args :],
[all …]
|
| /external/pytorch/torch/_inductor/kernel/ |
| D | mm_scaled.py | 261 input_nodes: Tuple[Any, ...]
264 input_nodes = (mat_a, mat_b, scale_a, scale_b)
268 input_nodes = (mat_a, mat_b, scale_a, scale_b, bias)
272 input_nodes, layout, out_dtype=out_dtype, use_fast_accum=use_fast_accum
290 input_nodes=input_nodes,
304 return autotune_select_algorithm("scaled_mm", choices, input_nodes, layout)
|
| D | mm.py | 178 input_nodes=(mat1, mat2),
195 input_nodes = [mat1, mat2]
209 input_nodes=(mat1, mat2),
223 input_nodes,
302 input_nodes=(mat1, mat2),
381 input_nodes=(inp_expanded, mat1, mat2),
585 input_nodes, argument
623 input_nodes=input_nodes,
703 input_nodes=(mat1, mat2),
713 input_nodes=(mat1, mat2),
[all …]
|
| /external/pytorch/torch/_inductor/codegen/ |
| D | cpp_template.py | 29 input_nodes, argument
35 self.input_nodes = input_nodes
58 unique(input_node.get_name() for input_node in self.input_nodes)
77 input_tensor_meta=TensorMeta.from_irnodes(self.input_nodes),
104 self.input_nodes,
|
| D | cpp_gemm_template.py | 241 input_nodes, argument
253 input_nodes,
263 _, k = input_nodes[0].get_size()
419 num_byte_A = get_num_byte(self.input_nodes[0].get_dtype())
420 num_byte_B = get_num_byte(self.input_nodes[1].get_dtype())
510 input_nodes, argument
519 input_indices = list(range(len(input_nodes)))
577 *maybe_to_dense(*reorder_and_filter(input_nodes, layout))
685 new_input_nodes, _ = reorder_and_filter(input_nodes, layout)
754 input_nodes=input_nodes,
[all …]
|
| /external/tensorflow/tensorflow/c/ |
| D | c_api_function.cc | 54 std::unordered_map<const Node*, std::vector<int>>* input_nodes) in ProcessInputs() argument
71 const auto& iter = input_nodes->find(node); in ProcessInputs()
72 if (iter == input_nodes->end()) { in ProcessInputs()
73 input_nodes->insert({node, {idx}}); in ProcessInputs()
113 const std::unordered_map<const Node*, std::vector<int>>& input_nodes, in ComputeBodyNodes() argument
118 const auto& iter = input_nodes.find(node); in ComputeBodyNodes()
119 if (iter == input_nodes.end()) { in ComputeBodyNodes()
163 std::unordered_map<const Node*, std::vector<int>> input_nodes; in TF_GraphToFunctionWithControlOutputs() local
165 &input_tensors, &input_nodes); in TF_GraphToFunctionWithControlOutputs()
195 fn_body, fn_name, num_opers, opers, input_nodes, &body_nodes); in TF_GraphToFunctionWithControlOutputs()
|
| /external/pytorch/torch/_inductor/autoheuristic/ |
| D | autoheuristic.py | 227 input_nodes: List[Any],
239 self.input_nodes = input_nodes
266 self.register_global_feedback(input_nodes, choices)
269 self, input_nodes: List[Any], choices: List[ChoiceCaller]
286 input_nodes: List[Any],
289 current_inputs_key = create_inputs_key(input_nodes)
300 inputs_key = create_inputs_key(input_nodes)
|
| /external/libaom/av1/encoder/x86/ |
| D | ml_avx2.c | 140 void av1_nn_predict_avx2(const float *input_nodes, in av1_nn_predict_avx2() argument
160 nn_propagate_input_multiple_of_8(input_nodes, layer_weights, layer_bias, in av1_nn_predict_avx2()
171 input_nodes, layer_weights, layer_bias, num_inputs_to_process, in av1_nn_predict_avx2()
183 av1_nn_propagate_4to8_sse3(&input_nodes[in], in av1_nn_predict_avx2()
199 av1_nn_propagate_4to4_sse3(&input_nodes[in], in av1_nn_predict_avx2()
210 av1_nn_propagate_4to1_sse3(&input_nodes[in], in av1_nn_predict_avx2()
223 __m128 input = _mm_load1_ps(&input_nodes[in_node]); in av1_nn_predict_avx2()
235 input_nodes = output_nodes; in av1_nn_predict_avx2()
|
| D | ml_sse3.c | 153 void av1_nn_predict_sse3(const float *input_nodes, in av1_nn_predict_sse3() argument
174 av1_nn_propagate_4to8_sse3(&input_nodes[in], in av1_nn_predict_sse3()
186 nn_propagate_8to4(&input_nodes[in], in av1_nn_predict_sse3()
197 av1_nn_propagate_4to4_sse3(&input_nodes[in], in av1_nn_predict_sse3()
208 nn_propagate_8to1(&input_nodes[in], in av1_nn_predict_sse3()
219 &input_nodes[in], &layer_weights[out * num_inputs + in], &total); in av1_nn_predict_sse3()
230 __m128 input = _mm_load1_ps(&input_nodes[in_node]); in av1_nn_predict_sse3()
239 input_nodes = output_nodes; in av1_nn_predict_sse3()
|
| /external/executorch/backends/arm/operators/ |
| D | op_add.py | 44 input_nodes = tutils.get_two_inputs(node)
48 for tensor in input_nodes
55 all(tensor.meta["val"].dtype == torch.int32 for tensor in input_nodes)
62 input_nodes, tosa_graph
|
| /external/libaom/av1/encoder/arm/ |
| D | ml_neon.c | 267 void av1_nn_predict_neon(const float *input_nodes, in av1_nn_predict_neon() argument
284 nn_propagate_4to8(num_inputs, input_nodes, in av1_nn_predict_neon()
290 nn_propagate_8to4(num_inputs, input_nodes, in av1_nn_predict_neon()
296 nn_propagate_4to4(num_inputs, input_nodes, in av1_nn_predict_neon()
302 nn_propagate_8to1(num_inputs, input_nodes, in av1_nn_predict_neon()
308 nn_propagate_4to1(num_inputs, input_nodes, in av1_nn_predict_neon()
314 nn_propagate_xto1(num_inputs, input_nodes, in av1_nn_predict_neon()
320 nn_propagate_xsto1(num_inputs, input_nodes, in av1_nn_predict_neon()
328 val += layer_weights[node * num_inputs + i] * input_nodes[i]; in av1_nn_predict_neon()
334 input_nodes = output_nodes; in av1_nn_predict_neon()
|
| /external/executorch/backends/qualcomm/quantizer/ |
| D | custom_annotation.py | 47 input_nodes = node.args[0]
49 first_input_node = input_nodes[0]
56 for input_node in input_nodes[1:]:
149 input_nodes = node.args[0]
150 assert isinstance(input_nodes, Sequence)
151 first_input_node = input_nodes[0]
159 for input_node in input_nodes[1:]:
|
| /external/pytorch/torch/fx/passes/utils/ |
| D | source_matcher_utils.py | 40 input_nodes: List[Node] = field(default_factory=list) variable in SourcePartition
98 input_nodes = set()
104 input_nodes.add(arg)
116 list(input_nodes),
|
| /external/executorch/backends/qualcomm/_passes/ |
| D | recompose_rms_norm.py | 40 input_len = len(src_partition.input_nodes)
42 input_node = src_partition.input_nodes[0]
44 inp_0, inp_1 = src_partition.input_nodes
|
| /external/tensorflow/tensorflow/core/tfrt/utils/ |
| D | graph_partition.cc | 76 absl::flat_hash_map<std::string, NodeInfo>& input_nodes, in PrepareSubgraphForFunctionConversion() argument
105 input_nodes.emplace(node->name(), node_info); in PrepareSubgraphForFunctionConversion()
216 const absl::flat_hash_map<std::string, NodeInfo>& input_nodes, in BuildPartitionedCallOp() argument
228 std::vector<DataType> input_dtypes(input_nodes.size()); in BuildPartitionedCallOp()
229 for (const auto& input_node : input_nodes) { in BuildPartitionedCallOp()
248 std::vector<NodeBuilder::NodeOut> call_node_inputs(input_nodes.size()); in BuildPartitionedCallOp()
249 for (const auto& input_node : input_nodes) { in BuildPartitionedCallOp()
458 absl::flat_hash_map<std::string, NodeInfo> input_nodes; in InsertTransferOps() local
470 inputs, outputs, host_device, func_name, input_nodes, output_nodes, in InsertTransferOps()
477 BuildPartitionedCallOp(func_name, host_device, device, input_nodes, in InsertTransferOps()
|
| /external/tensorflow/tensorflow/tools/graph_transforms/ |
| D | strip_unused_nodes.cc | 120 std::set<string> input_nodes; in StripUnusedNodes() local
123 input_nodes.insert(NodeNameFromInput(input)); in StripUnusedNodes()
141 if (input_nodes.count(current_input)) { in StripUnusedNodes()
169 if (input_nodes.count(node.name())) { in StripUnusedNodes()
|
| D | backports.cc | 34 [](const NodeMatch& match, const std::set<string>& input_nodes, in BackportConcatV2Transform()
71 const std::set<string>& input_nodes, in BackportTensorArrayV3Transform() argument
117 [](const NodeMatch& match, const std::set<string>& input_nodes, in BackportTensorArrayV3Transform() argument
|
| D | fold_constants_lib.cc | 54 StringPieceSet input_nodes; in ReplaceSendRecvs() local
61 input_nodes.insert(id.first); in ReplaceSendRecvs()
103 auto iter = input_nodes.find(new_node->name()); in ReplaceSendRecvs()
104 if (iter != input_nodes.end()) { in ReplaceSendRecvs()
105 input_nodes.erase(iter); in ReplaceSendRecvs()
111 for (StringPiece name : input_nodes) { in ReplaceSendRecvs()
|
| /external/tensorflow/tensorflow/core/tpu/graph_rewrite/ |
| D | configure_tpu_embedding_rewrite_pass.cc | 82 absl::Span<Node* const> input_nodes, in AddSetupPropagationEmbeddingNode() argument
88 AddNodeAttr("N", static_cast<int>(input_nodes.size()), &node_def); in AddSetupPropagationEmbeddingNode()
89 if (!input_nodes.empty()) { in AddSetupPropagationEmbeddingNode()
90 MergeDebugInfo(NodeDebugInfo(input_nodes[0]->def()), &node_def); in AddSetupPropagationEmbeddingNode()
96 for (int i = 0; i < input_nodes.size(); ++i) { in AddSetupPropagationEmbeddingNode()
97 graph->AddEdge(input_nodes[i], 0, *node, i); in AddSetupPropagationEmbeddingNode()
|