Load model from 1.5.1 and save/use it on 1.7.1

1

Hi,
I have an old model saved from PyTorch 1.5.1 using torch.save(model, "MyModel.pt").
When I try to use it on PyTorch 1.7.1, I get the following error:

torch.nn.modules.module.ModuleAttributeError: 'FrozenBatchNorm2d' object has no attribute 'eps'

Unfortunately, I don’t know the architecture used to create the model, so I cannot recreate the same in 1.7.1 and copy the weights.

What are my options?
Thank you in advance.

2

If you’ve stored the “complete” model, you would still need the model class definition and all source files, wouldn’t you, or did you use a pretrained model somehow?

3

It’s Mask R-CNN model from torchvision, but I know that I tuned many parameters, and I don’t remember which ones.

4

Where did you tune these parameters? Directly in the source file (in torchvision) or in another script?
I don’t think that torch.save(model, path) will store all source files with it, and would thus try to load them in torch.load.
Note that I’m speculating a lot here, as I’m not using this workflow, since it can break in many ways.

5

The parameters were mainly for the RPN.
I can load the old model in 1.7.1, but when I use it for inference I get the error I cited above. If the sources were store within the .pt file, I would likely not have this error.

6

I don’t think the sources are stored (which is the reason the loading could break) and, if I’m not mistaken, it’s a known pickle limitation.
Since you already can load the model, you could try to set the missing eps to 0, which seems to be the default based on this PR.

7

Thanks.

So if I understand what you suggest, what I have to do is:

  • Load the old model.
  • Find the FrozenBatchNorm2d layers
  • Create a FrozenBatchNorm2d layers with PyTorch 1.7.1 with the old eps parameter.
  • Copy the old layers weights/biases in the new layers
  • Replace the old layers by the new one.

Am I correct?
If yes, I have two questions:

  • How do I replace one layer by another one?
  • How do I copy the weights and bias form one layer into the new one?
8

I don’t think you have to replace the FrozenBatchNorm2d layers and could try to add the missing eps attribute directly to the already used layers.
If I understand your issue correctly, the error is only raised in the forward pass when the eps attribute is used, not while loading the state_dict. If so, then adding the eps before executing the forward pass might work.

9

Thanks.
But if the model is already created with a missing parameter/argument, how can I add it?

10

You can add attributes by directly assigning them:

model.frozen_bn1.eps = 0.
1 Like
11

Thanks, I’ll try it.

12

It worked, thanks a lot.

from torchvision.ops import misc
model = torch.load("...")
for name, layer in model.named_modules():
   if isinstance(layer, misc.FrozenBatchNorm2d):
       layer.eps = 0.
torch.save(model, "...")
1 Like