fix issue #278 in relax.op.masked_fill

[sbwww]

adding dtype argument when calling _ffi_api.full_like

[MasterJH5574]

Here it’s better to use the Python interface of full_like than the FFI interface, because the Python interface has clear semantics on the dtype, and the dtype can be None, which means being the same as the input dtype.

relax/python/tvm/relax/op/create.py

Lines 56 to 79 in 8ee979d

	def full_like(x: Expr, fill_value: Expr, dtype: Optional[Union[str, DataType]] = None) -> Expr:
	"""Construct a tensor such that
	- its shape is the same as the input data tensor's shape,
	- its value is filled with the input scalar fill value.
	Parameters
	----------
	x : relax.Expr
	The input tensor, which provides the shape, and dtype
	when the `dtype` field is not specified.
	fill_value : relax.Expr
	The value to fill. Must be a scalar tensor.
	dtype : Optional[Union[str, DataType]]
	The data type of the created tensor.
	If dtype is not given, it will by default use the dtype of the input tensor.
	Returns
	-------
	result : relax.Expr
	The result tensor.
	"""
	return _ffi_api.full_like(x, fill_value, dtype) # type: ignore

So here we can do from .create import full_like, and use values = full_like(x, value)

[sbwww]

Thanks for your suggestion, please check the update.

[MasterJH5574]

Thanks @sbwww for the fix! Just merged the corresponding PR in apache/tvm. We have a routine-based synchronization with TVM Unity branch, and will bring that commit to this repo in the next round of sync. So please allow me to close this PR :-)

[MasterJH5574]

It has now been brought in.