Description
Combined test coverage of src/ is currently 43% (measured across both backends with --cov-append), well below the 80% target. The fail_under gate is set low provisionally so CI can go green; this issue tracks backfilling tests and raising the gate to 80%.
Note: 43% is a floor — it was measured with ~121 tests failing (CUDA/backend-gating issues). Re-measure once the suite is green before setting intermediate targets.
Biggest gaps
| Module |
Coverage |
Statements |
core/torch/fit_compress.py |
0% |
474 |
core/torch/distillers.py |
0% |
315 |
core/torch/pruning_methods/fitcompress.py |
0% |
33 |
core/utils.py |
0% |
41 |
core/torch/train.py |
6% |
38 |
core/torch/optimizers.py |
14% |
45 |
core/torch/layers.py |
48% |
1244 |
fit_compress.py + distillers.py alone are ~790 untested statements (~8.5% of the codebase).
Tasks
Acceptance Criteria
Description
Combined test coverage of
src/is currently 43% (measured across both backends with--cov-append), well below the 80% target. Thefail_undergate is set low provisionally so CI can go green; this issue tracks backfilling tests and raising the gate to 80%.Note: 43% is a floor — it was measured with ~121 tests failing (CUDA/backend-gating issues). Re-measure once the suite is green before setting intermediate targets.
Biggest gaps
core/torch/fit_compress.pycore/torch/distillers.pycore/torch/pruning_methods/fitcompress.pycore/utils.pycore/torch/train.pycore/torch/optimizers.pycore/torch/layers.pyfit_compress.py+distillers.pyalone are ~790 untested statements (~8.5% of the codebase).Tasks
fit_compress.py,pruning_methods/fitcompress.py).distillers.py.core/utils.py,core/torch/train.py,core/torch/optimizers.py.fail_underinpyproject.tomlincrementally (e.g. 60 → 80) as coverage lands.Acceptance Criteria
[tool.coverage.report] fail_under = 80and the CI coverage job passes.