VGG19 RuntimeError: stride should not be zero

mps
1

I am using transfer learning to tune a pretrained VGG19 model. I am using MPS on Mac M1. Here is the printed model and error that occurs during forward pass:
MODEL:
VGG(
(features): Sequential(
(0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(1): ReLU(inplace=True)
(2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(3): ReLU(inplace=True)
(4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(6): ReLU(inplace=True)
(7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(8): ReLU(inplace=True)
(9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(11): ReLU(inplace=True)
(12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(13): ReLU(inplace=True)
(14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(15): ReLU(inplace=True)
(16): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(17): ReLU(inplace=True)
(18): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(19): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(20): ReLU(inplace=True)
(21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(22): ReLU(inplace=True)
(23): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(24): ReLU(inplace=True)
(25): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(26): ReLU(inplace=True)
(27): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(29): ReLU(inplace=True)
(30): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(31): ReLU(inplace=True)
(32): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(33): ReLU(inplace=True)
(34): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(35): ReLU(inplace=True)
(36): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)
(avgpool): AdaptiveAvgPool2d(output_size=(7, 7))
(classifier): Sequential(
(0): Linear(in_features=25088, out_features=8192, bias=True)
(1): ReLU(inplace=True)
(2): Linear(in_features=8192, out_features=256, bias=True)
(3): ReLU(inplace=True)
(4): Linear(in_features=256, out_features=2, bias=True)
)
)

ERROR:
Traceback (most recent call last):
File “~/Desktop/ObjectSecurity/adversarial-segmentation/catsdogs_torch.py”, line 124, in
pred = model(x)
File “~/miniforge3/envs/myenv2/lib/python3.9/site-packages/torch/nn/modules/module.py”, line 1186, in _call_impl
return forward_call(*input, **kwargs)
File “~/miniforge3/envs/myenv2/lib/python3.9/site-packages/torchvision/models/vgg.py”, line 67, in forward
x = self.avgpool(x)
File “~/miniforge3/envs/myenv2/lib/python3.9/site-packages/torch/nn/modules/module.py”, line 1186, in _call_impl
return forward_call(*input, **kwargs)
File “~/miniforge3/envs/myenv2/lib/python3.9/site-packages/torch/nn/modules/pooling.py”, line 1183, in forward
return F.adaptive_avg_pool2d(input, self.output_size)
File “~/miniforge3/envs/myenv2/lib/python3.9/site-packages/torch/nn/functional.py”, line 1214, in adaptive_avg_pool2d
return torch._C._nn.adaptive_avg_pool2d(input, _output_size)
RuntimeError: stride should not be zero

2

Could you check what the input shape to the adaptive avg pool layer is? My first guess is that could be the cause of the inferred 0 stride.

3

input shape to avgpool: torch.Size([64, 512, 4, 4])

4

Interesting, do you see any errors with the following snippet?

import torch

avgpool = torch.nn.AdaptiveAvgPool2d((7, 7))
a = torch.randn(64, 512, 4, 4, device='mps')
avgpool(a)
5

I get the same error. Any idea why? Should the input size be bigger than the output size?

6

I believe this is an issue that is specific to the MPS backend as I cannot observe it on cpu or cuda. Will open a bug to track this issue.

EDIT: issue is here [MPS] AdaptiveAvgPool2D doesn't accept input spatial size smaller than output shape · Issue #80732 · pytorch/pytorch · GitHub

1 Like