Commit 26ec25d

committed

fix simplenet for timm finetuning recipe

if during model creation, the number of classes differed from imagenets 1000 classes simplenet would crash because the pretrained weights wouldnt match. now when generating architectures, we check for this and if encountered with one, only swap the classifier after the model has been loaded with original weights.

1 parent 665f27b commit 26ec25dCopy full SHA for 26ec25d

File tree

1 file changed

-1

lines changed

ImageNet/training_scripts/imagenet_training/timm/models
- simplenet.py

1 file changed

-1

lines changed

`‎ImageNet/training_scripts/imagenet_training/timm/models/simplenet.py`

Lines changed: 1 addition & 1 deletion

Original file line number	Diff line number	Diff line change
`@@ -344,7 +344,7 @@ def forward_features(self, x: torch.Tensor) -> torch.Tensor:`
`344`	`344`	`return self.features(x)`
`345`	`345`
`346`	`346`	`def forward_head(self, x: torch.Tensor, pre_logits: bool = False):`
`347`		`- x_ = self.forward_features(x)`
	`347`	`+ x = self.forward_features(x)`
`348`	`348`	`if pre_logits:`
`349`	`349`	`return x`
`350`	`350`	`else:`

0 commit comments

Comments

(0)

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Commit 26ec25d

File tree

1 file changed

1 file changed

`‎ImageNet/training_scripts/imagenet_training/timm/models/simplenet.py`

0 commit comments