Operator

class webdnn.graph.operator.Operator(name=None)

Operator a.k.a layer or function in DNN computation graph.

Parameters:name (str) – the name. If None, automatically generated name is used.

append_input(name, var)

Append input variable

Parameters:

• name (str) – the name of input variable
• var (Variable) – the variable

append_output(name, var)

Append output variable

Parameters:

• name (str) – the name of output variable
• var (Variable) – the variable

copy()

Create new operator instance, which is initialized with parameters same as this operator.

inputs

input variables

outputs

output variables

remove_all()

Remove all input and output variables

remove_input(var)

Remove input variable

Parameters:var (Variable) – the variable

remove_output(var)

Remove output variable

Parameters:var (Variable) – the variable

replace(op_new)

Replace this operator by new operator. all variables connected with this operator will be disconnected and connected to the new operator.

Parameters:op_new (Operator) – the new operator

replace_input(v_old, v_new)

Replace input variable with other variable

Parameters:

• v_old (Variable) – the variable which is removed
• v_new (Variable) – the variable which is appended
• with_assert (bool) – If True, it is checked whether shape and order is same as variable which will be removed

replace_output(v_old, v_new)

Replace output variable with other variable

Parameters:

• v_old (Variable) – the variable which is removed
• v_new (Variable) – the variable which is appended
• with_assert (bool) – If True, it is checked whether shape and order is same as variable which will be removed