input = node.input[i]                if input in id_dict:                    node.input[i] = id_dict[input]      def generate_edges(self):        for node in self.nodes:            for input in node.input:                if input in self.edges:                    self.edges[input].append(node)                else:                    self.edges[input] = [node]      @staticmethod    def get_scope_name(name):        index = name.rfind('/')        if index == -1:            return ""        return name[0:index]      def in_conv2d_scope(self, name):        inner_scope = TFConverter.get_scope_name(name)        if inner_scope == "":            return False;        for scope in self.conv2d_scope_names:            index = inner_scope.find(scope)            if index == 0:                return True        return False      def in_dense_scope(self, name):        inner_scope = TFConverter.get_scope_name(name)        if inner_scope == "":            return False;        for scope in self.dense_scope_names:            index = inner_scope.find(scope)            if index == 0:                return True        return False     def generate_sub_block_op_scope_info(self):        # mostly, conv2d/dense is a sub block in graph, get the scope name        for node in self.nodes:            if node.op == 'Conv2D':                scope = TFConverter.get_scope_name(node.name)                # for the case tf.nn.conv2d is called directly                if scope == '':                    continue                # for the case tf.nn.conv2d is called within a scope                if scope + '/kernel' not in self.name_node_dict:                    continue                self.conv2d_scope_names.add(scope)            elif node.op == 'MatMul':                scope = TFConverter.get_scope_name(node.name)                # for the case tf.nn.dense is called directly                if scope == '':                    continue                # for the case tf.nn.dense is called within a scope                if scope + '/kernel' not in self.name_node_dict and scope.split('/Tensordot')[0] + '/kernel' not in self.name_node_dict:                    continue                self.dense_scope_names.add(scope.split('/Tensordot')[0])         # get the input name to the conv2d/dense sub block        for node in self.nodes:            scope = TFConverter.get_scope_name(node.name)            if scope in self.conv2d_scope_names:                if node.op == 'Conv2D' or node.op == 'Shape':                    for inp in node.input:                        if TFConverter.get_scope_name(inp) != scope:                            self.conv2d_scopename_inputname_dict[scope] = inp            elif scope in self.dense_scope_names:                if node.op == 'MatMul' or node.op == 'Shape':                    for inp in node.input:                        if TFConverter.get_scope_name(inp) != scope:                            self.dense_scopename_inputname_dict[scope] = inp            elif scope.split('/Tensordot')[0] in self.dense_scope_names:                if node.op == 'Transpose':                    for inp in node.input:                        if TFConverter.get_scope_name(inp).find(scope)<0 and TFConverter.get_scope_name(inp).find(scope.split('/')[0])<0:                            self.dense_scopename_inputname_dict[scope.split('/Tensordot')[0]] = inp      def run(self):        self.generate_name_node_dict()        self.generate_output_names()        self.remove_identity()        self.generate_edges()        self.generate_sub_block_op_scope_info()         if self.dump4tb:            self.dump_for_tensorboard()         self.dump_to_file()  def convert_from_tensorflow(infile, outfile, dump4tb):    with open(infile, 'rb') as f:        # read the file in .proto format        graph_def = tf.GraphDef()        graph_def.ParseFromString(f.read())        nodes = graph_def.node     converter = TFConverter(graph_def, nodes, outfile, dump4tb)    converter.run()

Lines 501–607 of 607. Previous lines