float default values for Scalar in native_functions.yaml don't work #8484
Description
rrelu is affected by this and (probably) gives incorrect behavior (I have not verified this)
func: rrelu(Tensor self, Scalar lower=0.125, Scalar upper=0.3333333333333333, bool training=false, Generator* generator=nullptr) -> Tensor
variants: function
func: rrelu_(Tensor self, Scalar lower=0.125, Scalar upper=0.3333333333333333, bool training=false, Generator* generator=nullptr) -> Tensor
variants: function
I gdb'ed into rrelu and printed out "lower":
(gdb) p upper
$1 = {tag = at::Scalar::Tag::HAS_i, v = {d = 0, i = 0}, t = {pImpl =
0x7fffe951f120 <at::UndefinedTensor::_singleton>}}
(gdb) p lower
$2 = {tag = at::Scalar::Tag::HAS_i, v = {d = 0, i = 0}, t = {
pImpl = 0x7fffe951f120 <at::UndefinedTensor::_singleton>}}
they have the at::Scalar::Tag::HAS_i flag, so they're both being treated as integers. Sending them to std::cout gives 0.