crutcher / bimm / #173

72%

push

feat: enhance ResNet API and resnet_tiny example with DropBlock and stochastic depth support

- Added new API methods to ResNet for setting DropBlock options and stochastic depth configurations (`with_standard_drop_block_prob`, `with_stochastic_depth_drop_path_rate`).
- Introduced `DropBlockOptions` support across `ResidualBlock`, `LayerBlock`, and ResNet layers.
- Updated `resnet_tiny` example to use ResNet-101 and new API methods for configurable DropBlock probability and drop path rate.
- Replaced `AdamW` optimizer with `SGD` and added gradient clipping and weight decay configurations for better training control.
- Modified default training settings: reduced batch size to 32, adjusted learning rate, and increased epochs to 100.

0 of 49 new or added lines in 3 files covered. (0.0%)

2255 of 2925 relevant lines covered (77.09%)

39.77 hits per line