TypeError: 'Proxy' object cannot be interpreted as an integer

[MHGL]

🐛 Bug

I get an error

• loop in forward and loop count num from input x
• call torch.quantization.quantize_fx.prepare_fx

To Reproduce

Steps to reproduce the behavior:

1. code example

import torch
from torch.quantization import get_default_qconfig
from torch.quantization.quantize_fx import prepare_fx

# init module
class MyModule(torch.nn.Module):
    def __init__(self):
        super(MyModule, self).__init__()
        ...

    def forward(self, x):
        for i in range(x.size(1)):
            x += 1
        return 

torch_model = MyModule().eval()

# fx
s_qconfig_dict = {'': get_default_qconfig("fbgemm")}
prepare_fx(torch_model, s_qconfig_dict)

2. stack traces

Traceback (most recent call last):
  File "mini_code.py", line 22, in <module>
    prepare_fx(torch_model, s_qconfig_dict)
  File "/opt/conda/lib/python3.8/site-packages/torch/quantization/quantize_fx.py", line 392, in prepare_fx
    return _prepare_fx(model, qconfig_dict, prepare_custom_config_dict)
  File "/opt/conda/lib/python3.8/site-packages/torch/quantization/quantize_fx.py", line 174, in _prepare_fx
    graph_module = GraphModule(model, tracer.trace(model))
  File "/opt/conda/lib/python3.8/site-packages/torch/fx/symbolic_trace.py", line 571, in trace
    self.create_node('output', 'output', (self.create_arg(fn(*args)),), {},
  File "mini_code.py", line 14, in forward
    for i in range(x.size(1)):
TypeError: 'Proxy' object cannot be interpreted as an integer

Expected behavior

Environment

• PyTorch Version: 1.9.0
• OS (e.g., MacOS, Linux): Ubuntu20.04 LTS
• How you install python (anaconda, virtualenv, system): miniconda
• python version (e.g. 3.7): 3.8.5
• any other relevant information:
  • gpu: GeForce GTX 1650
  • driver: Driver Version: 460.80
  • CUDA: CUDA Version: 11.2

[xiaopengaia]

How to solve it? I also met it ..... Thanks!

[xiaopengaia]

I get it, Thank you!

[PHaiJun]

I got a same error, how to solve it?

[frothmoon]

@xiaopengaia How to solve it? I also met it ..... Thanks!