Skip to content

Change the classifier input from 2048 to 1000. #5232

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Jan 5, 2021
Merged

Change the classifier input from 2048 to 1000. #5232

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
e6bfb0c
Change the classifier input from 2048 to 1000.
LaserBit Dec 22, 2020
e692b40
Update docs for Imagenet example
LaserBit Dec 26, 2020
06c2df8
Apply suggestions from code review
rohitgr7 Jan 4, 2021
8ab059d
Merge branch 'master' into master
SkafteNicki Jan 5, 2021
d48275f
Merge branch 'master' into master
SkafteNicki Jan 5, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 11 additions & 5 deletions docs/source/transfer_learning.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,16 +52,22 @@ Example: Imagenet (computer Vision)

class ImagenetTransferLearning(LightningModule):
def __init__(self):
super().__init__()

# init a pretrained resnet
num_target_classes = 10
self.feature_extractor = models.resnet50(pretrained=True)
self.feature_extractor.eval()
backbone = models.resnet50(pretrained=True)
num_filters = backbone.fc.in_features
layers = list(backbone.children())[:-1]
self.feature_extractor = torch.nn.Sequential(*layers)

# use the pretrained model to classify cifar-10 (10 image classes)
self.classifier = nn.Linear(2048, num_target_classes)
num_target_classes = 10
self.classifier = nn.Linear(num_filters, num_target_classes)

def forward(self, x):
representations = self.feature_extractor(x)
self.feature_extractor.eval()
with torch.no_grad():
representations = self.feature_extractor(x).flatten(1)
x = self.classifier(representations)
...

Expand Down