Bug? torch.jit.script(model) loses list attributes with mixed element types

I found that the scripted model (by torch.jit.script) silently loses list attributes where the elements have mixed types, e.g. the following code snippet:

import torch
from torch import nn

class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.a = [1, 2]
        self.b = [1, 2.2]

m = Model()
print(m.a)  # [1, 2]
print(m.b)  # [1, 2.2]

sm = torch.jit.script(m)
print(sm.a)  # [1, 2]
print(sm.b)  # error, no attribute b, why?

The attribute a is a list of int and the scripting worked fine, but b (with mixed element types, int and double) is silently lost. The scripted model sm doesn’t have it anymore. I think the biggest concern is that the scripting process didn’t give any error and just silently returned a scripted model.

Is this a known bug (or WAI) ?

torch version: 1.12.0