Checklist USA-NA-AIO 2026 Round 1 · Task 8
ResNet-50 Parameters and a Frozen Backbone
Count parameters and output shapes of a pretrained ResNet-50 and build a frozen truncated backbone for 5-class classification.
The task
Problem 8 (25 points) loads torchvision's pretrained resnet50. The parts ask for the total number of learnable parameters, the output shape of model.layer2 for input (B, 3, 224, 224), the number of learnable parameters in the convolutional modules of model.layer3[2] without using numel, and a 5-class classifier that uses the network truncated after model.layer3[4] as a frozen backbone.
Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.
At a glance
- You submit
- Code and numeric answers.
- Rules
- Part 8.3 forbids numel on Parameter objects.
- Format
- 2026 USA-NA-AIO Round 1, 30 January 2026 (date printed on the problem set). Individual, proctored.