pytorch jit overload with inheritance

test.py

import torch
from torch.nn.utils.rnn import PackedSequence
from typing import overload, Optional

class Base(torch.nn.Module):
    def __init__(self):
        super().__init__()
    @overload
    @torch._jit_internal._overload_method
    def forward(self, inputs, hx=None):
        # type: (PackedSequence, Optional[Tensor]) -> PackedSequence
        pass
    @overload
    @torch._jit_internal._overload_method
    def forward(self, inputs, hx=None):
        # type: (Tensor, Optional[Tensor]) -> Tensor
        pass
    def forward(self, inputs, hx=None):
        return inputs
class Derive(Base):
    pass


class Derive2(Base):
    @overload
    @torch._jit_internal._overload_method
    def forward(self, inputs, hx=None):
        # type: (PackedSequence, Optional[Tensor]) -> PackedSequence
        pass
    @overload
    @torch._jit_internal._overload_method
    def forward(self, inputs, hx=None):
        # type: (Tensor, Optional[Tensor]) -> Tensor
        pass
    def forward(self, inputs, hx=None):
        return Base.forward(self, inputs, hx)

torch.jit.script(Base())

try:
    torch.jit.script(Derive())#this doesn't work
except Exception as e:
    print(e)
try:
    torch.jit.script(Derive2())#doesn't work either
except Exception as e:
    print(e)