Tensorflow Frontend fill op error


#1

Fill op directly uses inputs[0] as an extra argument, and it does not work for me. Did anyone test this op before?
def _fill():
def _impl(inputs, attr, params):
fill_arg = params.pop(inputs.pop(1).list_output_names()[0])
new_inputs = []
return AttrCvt(
op_name=‘full’,
extras={‘shape’:inputs[0],
‘fill_value’:fill_arg.asnumpy()[0], ‘dtype’:attr[‘T’].name},
ignores=[‘index_type’, ‘T’])(new_inputs, attr)
return _impl


#2

inputs is a list of symbols. If inputs[0] is available in params, try extracting the value from params and construct the argument.


#3

Thank you, but inputs[0] is not in params, can I open an issue for this problem?


#4

not in param means the argument depends on the input data which is not available while compilation. This situation is not possible with NNVM.