ONNX frontend bug with MissingOperator
victorromeo opened this issue · comments
victorromeo commented
In #133 Missing Operator was found to be incorrectly using dict key, due to missing Namespace term.
utensor_cgen/backend/utensor/code_generator/rearch/_operators/_impls.py
class _MissingOperator(_Operator):
op_type = "_MissingOperator"
def get_declare_snippet(self, op_var_name, with_const_params=True):
return None
def get_eval_snippet(self, op_var_name, op_info, tensor_var_map):
return MissingOpEvalSnippet(op_info, op_var_name, tensor_var_map)
def get_construct_snippet(self, op_var_name):
return None
OperatorFactory._operators[_MissingOperator.op_type] = _MissingOperator # <== key should be a 2 term tensor, (namespace, op_type)
dboyliao commented
That was on purpose, since _MissinOperator
is a special operator which is used for printing out information only.
Anyway, it's fixed on develop
.
Closing issue.