crutcher / bimm / #172

72%

push

refactor: update `resnet_tiny` defaults and improve ResNet API

- Increased `batch_size` default from 32 to 512 for enhanced performance with larger workloads.
- Adjusted `learning_rate` default from `1.0e-4` to `1.0e-3` for faster convergence.
- Replaced hardcoded image shape with a reusable `ImageShape` struct.
- Deprecated gradient clipping in optimizer configuration for simplicity.
- Added new `resnet34` constructor and corresponding block configuration in ResNet API.
- Enhanced `LayerBlock` tests with `build` validation and resolution checks.
- Fixed typos in activation layer enum (`GeLu` -> `Gelu`) and updated naming consistency.

3 of 9 new or added lines in 3 files covered. (33.33%)

2254 of 2876 relevant lines covered (78.37%)

40.45 hits per line