# -*- coding:UTF-8 -*- # Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License" # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. import copy import pandas as pd from x2paddle.optimizer.code_optimizer.layer_code_generator import rename_layers def construct_attrs_table(sub_layers_list, node_name2sub_layers): """ 构造不同属性的表格。 """ def get_node_name(sub_layers): for k, v in node_name2sub_layers.items(): if v == sub_layers: node_name = k break return node_name sub_layers = sub_layers_list[0] _, _, new_names = rename_layers(sub_layers) table = list() node_names = list() for sub_layers in sub_layers_list: attrs = dict() node_names.append(get_node_name(sub_layers)) for i, (layer_id, layer) in enumerate(sub_layers.items()): for k, v in layer.attrs.items(): attrs[new_names[i] + "_{}".format(k)] = v table.append(attrs) pd_table = pd.DataFrame(table, index=node_names) return pd_table def get_inputs_outputs(pd_graph, layers): inputs = list() outputs = list() cur_outputs = list() layer_ids = list(layers.keys()) for layer_id, layer in layers.items(): # 获取输出节点名字 if layer_id not in pd_graph.edges_out: for index, output_name in enumerate(layer.outputs): if not output_name.startswith("x") or output_name in outputs \ or layer.kernel == "prim.assert": continue elif layer.kernel == "prim.if" or layer.kernel == "prim.loop": if index != 0: outputs.append(output_name) elif output_name not in outputs: outputs.append(output_name) else: for out_layer_id in pd_graph.edges_out[layer_id]: if out_layer_id not in layer_ids: for index, output_name in enumerate(layer.outputs): if not output_name.startswith("x") or output_name in outputs \ or layer.kernel == "prim.assert": continue elif layer.kernel == "prim.if" or layer.kernel == "prim.loop": if index != 0: outputs.append(output_name) else: outputs.append(output_name) # 获取输入节点名字 for k, v in layer.inputs.items(): if v not in cur_outputs and v not in inputs: inputs.append(v) if ("paddle.nn" in layer.kernel and "functional" not in layer.kernel): cur_outputs.extend(layer.outputs[1:]) else: cur_outputs.extend(layer.outputs) return inputs, outputs def get_inputs_count(pd_graph, sub_layers): input_ct2sub_layer_id = dict() for i, sub_layer in enumerate(sub_layers): inputs, outputs = get_inputs_outputs(pd_graph, sub_layer) if len(inputs) not in input_ct2sub_layer_id: input_ct2sub_layer_id[len(inputs)] = [i] else: input_ct2sub_layer_id[len(inputs)].append(i) return input_ct2sub_layer_id def distinguish_sequential(pd_graph, module_name, sub_layers, sub_identifiers, node_name2sub_layers): """ 获取不同的layers组成的序列 """ def distinguish_sequential_by_inputs(part_layers, part_identifiers, part_module_name): new_sub_layers = dict() new_sub_sequentials = dict() sequentials2attrs_table = dict() input_ct2sub_layer_id = get_inputs_count(pd_graph, part_layers) if len(input_ct2sub_layer_id) == 1: new_sub_layers["{}".format(part_module_name)] = part_layers new_sub_sequentials["{}".format(part_module_name)] = part_identifiers sequentials2attrs_table["{}".format(part_module_name)] = construct_attrs_table(part_layers, node_name2sub_layers) else: for i, (k, indexes) in enumerate(input_ct2sub_layer_id.items()): new_sub_layers["{}__{}".format(part_module_name, i)] = list() new_sub_sequentials["{}__{}".format(part_module_name, i)] = list() for index in indexes: new_sub_layers["{}__{}".format(part_module_name, i)].append(part_layers[index]) new_sub_sequentials["{}__{}".format(part_module_name, i)].append(part_identifiers[index]) sequentials2attrs_table["{}__{}".format(part_module_name, i)] = \ construct_attrs_table(new_sub_layers["{}__{}".format(part_module_name, i)], node_name2sub_layers) return new_sub_layers, new_sub_sequentials, sequentials2attrs_table new_sub_layers = dict() new_sub_sequentials = dict() sequentials2attrs_table = dict() identifiers_str_list = list() for identifiers in sub_identifiers: identifiers_str_list.append(", ".join(list(identifiers.values()))) identifiers_str_set = list(set(identifiers_str_list)) if len(identifiers_str_set) == 1: return distinguish_sequential_by_inputs(sub_layers, sub_identifiers, module_name) else: for i in range(len(identifiers_str_set)): new_sub_layers["{}{}".format(module_name, i)] = list() new_sub_sequentials["{}{}".format(module_name, i)] = list() no_same_module_count = 0 for j, identifiers in enumerate(sub_identifiers): identifiers_str = identifiers_str_list[j] for i in range(len(identifiers_str_set)): if identifiers_str_set[i] == identifiers_str: is_diff = False if identifiers_str_set[i].replace(", ", "").isdigit() or module_name == "ModuleList": new_sub_layers["{}{}".format(module_name, len(identifiers_str_set) + no_same_module_count)] = [sub_layers[j]] new_sub_sequentials["{}{}".format(module_name, len(identifiers_str_set) + no_same_module_count)] = [identifiers] no_same_module_count += 1 else: new_sub_layers["{}{}".format(module_name, i)].append(sub_layers[j]) new_sub_sequentials["{}{}".format(module_name, i)].append(identifiers) break new_new_sub_layers = dict() new_new_sub_sequentials = dict() for k, v in new_sub_layers.items(): part_sub_layers, part_sub_sequentials, part_sequentials2attrs_table = \ distinguish_sequential_by_inputs(v, new_sub_sequentials[k], k) new_new_sub_layers.update(part_sub_layers) new_new_sub_sequentials.update(part_sub_sequentials) sequentials2attrs_table.update(part_sequentials2attrs_table) return new_new_sub_layers, new_new_sub_sequentials, sequentials2attrs_table