Torch.fx.symbolic_trace doesn't preserve module methods

Are module methods other than forward preserved? I expected scale to be there in the traced_mod.

import torch
import torch.fx

class MyModule(torch.nn.Module):
    def __init__(self):
        self.linear = torch.nn.Linear(512, 512)

    def forward(self, x):
        x = self.linear(x)
        x = torch.relu(x)
        return x

    def scale(self, x):
        return x * 2

mod = MyModule()

traced_mod = torch.fx.symbolic_trace(mod)

x = torch.randn(1, 512)

orig_out = mod.scale(x)

traced_out = traced_mod.scale(x)
Traceback (most recent call last):
  File "", line 25, in <module>
    traced_out = traced_mod.scale(x)
  File "~/venv_sys/lib/python3.8/site-packages/torch/nn/modules/", line 1177, in __getattr__
    raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'MyModule' object has no attribute 'scale'

PyTorch version:

python -c "import torch; print(torch.__version__)"

Fx only trace the module about forward function. Do not trace your function you write.