Module 'Sequential' has no attribute '_modules' :

[SerBad]

🐛 Describe the bug

import torch
import torch.nn as nn
class MyModule(nn.Module):
    def __init__(self):

        super(MyModule, self).__init__()

        self.seq = nn.Sequential(
            nn.Linear(10, 20),
            nn.ReLU(),
            nn.Linear(20, 5)
        )

def forward(self, x):
    return self.seq(x)

my_module = MyModule()
scripted_module = torch.jit.script(my_module)

RuntimeError: 
Module 'Sequential' has no attribute '_modules' :
  File "/home/zhou/.local/lib/python3.10/site-packages/torch/nn/modules/container.py", line 251
    @property
    def modules(self):
        return self._modules
               ~~~~~~~~~~~~~ <--- HERE

Versions

version = '2.0.0+cu117'
debug = False
cuda = '11.7'
git_version = 'c263bd43e8e8502d4726643bc6fd046f0130ac0e'
hip = None

cc @EikanWang @jgong5 @wenzhe-nrv @sanchitintel

[pat749]

When I am running this code , I am not seeing any error. Here is the collab link .

I think there is error in your system locally, Here is also alternative way to fix this error is to properly indent the forward method within the MyModule class.

import torch
import torch.nn as nn

class MyModule(nn.Module):
    def __init__(self):
        super(MyModule, self).__init__()
        self.seq = nn.Sequential(
            nn.Linear(10, 20),
            nn.ReLU(),
            nn.Linear(20, 5)
        )

    def forward(self, x):
        return self.seq(x)

my_module = MyModule()
scripted_module = torch.jit.script(my_module)

Here is the implementation of this code also in google colab.

[SerBad]

do you use torch 2.0.0 ? I use torch 2.0.0.
If you use torch 1.13.0, it works.
but torch 2.0.0 did't work.

[pat749]

Yes I used torch 2.0.0 version . It works on torch 2.0.0.
Thank you

[SerBad]

It works when I remove @Property on 'def modules(self):',
and I don't know why my version have @Property.
maybe because I compile pytorch on local system? I don't know the reason.