ControlNet / tensorneko / 21825120772

87%

master: 87%

Build Type

push

github

Committed by ControlNet

Commit Message

refactor: improve code quality - assert→raise, safe tensor ops, public API usage

P2 fixes across 26 source files and 8 test files:

- Replace ~25 assert statements with proper exceptions (ValueError, TypeError, RuntimeError)
  for user-facing validation in IO, preprocessing, callbacks, layers, and modules
- Replace .data direct tensor mutation with torch.no_grad()+copy_/mul_ in layers
- Replace nn.init via .data.uniform_() with torch.nn.init.uniform_() in VectorQuantizer
- Remove debug print() from MaskedConv2d
- Add g_steps/d_steps validation in GAN.__init__
- Switch callbacks from trainer._logger_connector.callback_metrics to trainer.callback_metrics
- Add __len__ to NestedDataset
- Fix writer error messages (reader→writer)
- Update all affected tests to expect new exception types

Run Details

118 of 156 new or added lines in 26 files covered. (75.64%)

6 existing lines in 3 files now uncovered.

5816 of 6670 relevant lines covered (87.2%)

0.87 hits per line