Error: loading UNet xse_resnext50

[peterdudfield]

Get an error when loading an UNet xse_resnext50.

Please confirm you have the latest versions of fastai, fastcore, and nbdev prior to reporting a bug (delete one): YES

Describe the bug
A clear and concise description of what the bug is.

To Reproduce

from fastai.vision.learner import create_unet_model
from fastai.vision.models.xresnet import xse_resnext50

kwargs = {
    "img_size": (128, 256),
    "n_in": 15,
    "n_out": 25,
    "pretrained": False,
}

model = create_unet_model(arch=xse_resnext50, **kwargs)

Expected behavior

No error and the model to load

Error with full stack trace

Traceback (most recent call last):
  File "/Users/peterdudfield/Documents/Github/nowcasting_forecast/scripts/error.py", line 16, in <module>
    model = create_unet_model(arch=xse_resnext50, **kwargs)
  File "/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/site-packages/fastai/vision/learner.py", line 243, in create_unet_model
    model = arch(pretrained)
  File "/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/site-packages/fastai/vision/models/xresnet.py", line 99, in xse_resnext50
    def xse_resnext50(n_out=1000, pretrained=False, **kwargs):  return XResNet(SEResNeXtBlock, 4, g1, n_out=n_out, **se_kwargs2, **kwargs)
  File "/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/site-packages/fastai/vision/models/xresnet.py", line 47, in __init__
    nn.Linear(block_szs[-1]*expansion, n_out),
  File "/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 97, in __init__
    self.weight = Parameter(torch.empty((out_features, in_features), **factory_kwargs))
TypeError: empty() received an invalid combination of arguments - got (tuple, dtype=NoneType, device=NoneType), but expected one of:
 * (tuple of ints size, *, tuple of names names, torch.memory_format memory_format, torch.dtype dtype, torch.layout layout, torch.device device, bool pin_memory, bool requires_grad)
 * (tuple of ints size, *, torch.memory_format memory_format, Tensor out, torch.dtype dtype, torch.layout layout, torch.device device, bool pin_memory, bool requires_grad)

Potential Solution
By changing the line to

model = arch(pretrained=pretrained)

I can do a PR for this if you like

[peterdudfield]

Feel like we should keep this issue open until it is fixed and released?

[jph00]

It's fixed in #3746 . Please let me know if it doesn't work for you.

(We close issues when they're fixed in the repo, not when the release occurs.)

[peterdudfield]

It's fixed in #3746 . Please let me know if it doesn't work for you.

(We close issues when they're fixed in the repo, not when the release occurs.)

Sorry i was / am a littel confused. #3746 says it was closed with out merging, so just trying to see where the commit is

[jph00]

Oh sorry my bad! Yes you're right this is that commit that I slid into the general nbdev2 fixes. You can use git blame to find the commit that fixed it if you're curious.

https://github.com/fastai/fastai/blob/master/fastai/vision/learner.py#L240

[peterdudfield]

Thanks @jph00