vmap fails to call torch.compiled function about pytorch HOT 2 OPEN

torch/_subclasses/fake_tensor.py:1153: in _cached_dispatch_impl
    output = self._dispatch_impl(func, types, args, kwargs)
torch/_subclasses/fake_tensor.py:1539: in _dispatch_impl
    (flat_args, flat_arg_fake_tensors) = self.validate_and_convert_non_fake_tensors(
torch/_subclasses/fake_tensor.py:1832: in validate_and_convert_non_fake_tensors
    validated_args = [validate(a) for a in flat_args]
torch/_subclasses/fake_tensor.py:1832: in <listcomp>
    validated_args = [validate(a) for a in flat_args]
torch/_subclasses/fake_tensor.py:1822: in validate
    raise AssertionError(
E   torch._dynamo.exc.BackendCompilerFailed: backend='aot_eager' raised:
E   AssertionError: Please convert all Tensors to FakeTensors first or instantiate FakeTensorMode with 'allow_non_fake_inputs'. Found in aten.sin.default(BatchedTensor(lvl=1, bdim=0, value=
E       FakeTensor(..., size=(3, 4))
E   ))
E
E   While executing %sin : [num_users=1] = call_method[target=sin](args = (%l_x_,), kwargs = {})
E   Original traceback:
E     File "/home/guilhermeleobas/git/pytorch/test/dynamo/test_higher_order_ops.py", line 5290, in wrapped_fn
E       return x.sin()
E