Skip to content

Pin librosa <0.11 to protect voicing-gate calibration; fix fork HQ lane test stub #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
KhryptorGraphics wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Pin librosa <0.11 to protect voicing-gate calibration; fix fork HQ lane test stub #2

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
762f0b7
fix: pin librosa <0.11 to protect voicing-gate calibration
claude Sep 7, 2026
a4b67eb
fix: accept method kwarg in fork HQ lane _extract_pitch stub
claude Sep 7, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion requirements-runtime.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,11 @@ torch>=2.0
torchaudio>=2.0
numpy>=1.24
scipy>=1.10
librosa>=0.10
# <0.11: pyin's voiced decision changed in 0.11 (white noise scores ~0.43
# voiced vs 0.0 on 0.10), which silently shifts the calibrated multi-speaker
# voicing gates (multi_speaker_merge_voiced_min et al). Re-calibrate before
# raising this bound.
librosa>=0.10,<0.11
soundfile>=0.12

# Web
Expand Down
4 changes: 2 additions & 2 deletions requirements.lock
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,8 +5,8 @@
"backend": {
"requirements_runtime": {
"path": "requirements-runtime.txt",
"sha256": "a2e2fc3c1448ddd6b1caf661daa56a6b37086c36aba586b4e3a2f29a0a916133",
"line_count": 31
"sha256": "c013d714111e1198048b2a47bbdf7a05b706ae5e5cb05f3686877ed957726f56",
"line_count": 32
},
"frontend_lock": {
"path": "frontend/package-lock.json",
Expand Down
4 changes: 4 additions & 0 deletions src/auto_voice/inference/singing_conversion_pipeline.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,10 @@ def _voiced_fraction(audio: np.ndarray, sample_rate: int, max_s: float = 20.0) -
(calibrated leads: 0.76-0.87); harmony stacks / textures it would butcher
score low (calibrated: 0.14-0.51). Measured on up to ``max_s`` of active
audio with pyin fmin=80 fmax=1000, matching the calibration convention.

Calibration is coupled to librosa 0.10 pyin semantics (0.11 changed the
voiced decision: white noise jumps from 0.0 to ~0.43) — hence the
``librosa<0.11`` bound in requirements-runtime.txt.
"""
import librosa

Expand Down
3 changes: 2 additions & 1 deletion tests/test_fork_hq_lane.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,8 @@ def infer(self, audio, spk, emb, sr):

monkeypatch.setattr(p, "_get_separator", lambda: FakeSep())
monkeypatch.setattr(p, "_get_model_manager", lambda: FakeMM())
monkeypatch.setattr(p, "_extract_pitch", lambda a, sr: np.zeros(8, np.float32))
monkeypatch.setattr(p, "_extract_pitch",
lambda a, sr, method=None: np.zeros(8, np.float32))
Comment on lines +36 to +37

@greptile-apps greptile-apps Bot Sep 7, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Pitch forwarding remains untested

The updated stub accepts but ignores method, and this test does not configure an f0_method. It will therefore still pass if the HQ lane stops forwarding a selected pitch-extraction method. Exercise a non-None override and assert the received argument so this routing contract is protected.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!


res = p._convert_song_fork_hq(str(song), "prof-1", 1.0, 0.9, True, "balanced", 0.0)

Expand Down
Loading