Drop DER scoring and the pyannote-metrics dependency#105
Merged
Conversation
The Homebrew bottle build failed for v0.1.1 and v0.1.2: pyannote-metrics (the `assembly eval --speaker-labels` DER scorer) drags in scikit-learn → scipy/numpy/pandas, and Homebrew's `--build-bottle` compiles every resource from source (`--no-binary :all:`). scipy has no Fortran/OpenBLAS toolchain on the macOS runner, so the bottle never built. DER was the only consumer of pyannote, so remove it wholesale: - delete aai_cli/der.py and route `assembly eval` to WER-only (drop --speaker-labels/--collar, speaker-turn dataset loading, and the reference-optional path that only existed for diarization-only datasets). - drop pyannote-metrics from pyproject + relock (removes pyannote-core/ -database, scikit-learn, scipy, numpy, pandas, joblib, threadpoolctl, sortedcontainers, six, python-dateutil, narwhals, tzdata, pyyaml). - prune those resources + the now-orphaned libyaml dep from the formula, and narrow the install probe to `import jiwer`. - update the eval/import-hygiene tests, snapshots, import-linter contracts, and README. WER scoring (jiwer) is unaffected.
cryptography is only pulled in by secretstorage, the Linux D-Bus keyring backend — macOS uses the Keychain backend, which needs neither. Move the cryptography resource and the openssl@3 link dependency into the on_linux block (alongside jeepney/secretstorage) so the arm64 bottle stops compiling cryptography from source (a Rust + OpenSSL build) for no benefit. cffi stays top-level: sounddevice needs it on every platform for audio capture.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The Homebrew bottle build failed for v0.1.1 and v0.1.2: pyannote-metrics
(the
assembly eval --speaker-labelsDER scorer) drags in scikit-learn →scipy/numpy/pandas, and Homebrew's
--build-bottlecompiles every resourcefrom source (
--no-binary :all:). scipy has no Fortran/OpenBLAS toolchain onthe macOS runner, so the bottle never built.
DER was the only consumer of pyannote, so remove it wholesale:
assembly evalto WER-only (drop--speaker-labels/--collar, speaker-turn dataset loading, and the
reference-optional path that only existed for diarization-only datasets).
-database, scikit-learn, scipy, numpy, pandas, joblib, threadpoolctl,
sortedcontainers, six, python-dateutil, narwhals, tzdata, pyyaml).
and narrow the install probe to
import jiwer.and README.
WER scoring (jiwer) is unaffected.