Algorithm info: 'bs_polarformer': InferenceModel(
checkpoint=HuggingFaceFile(repo_id='hunterhogan/bs_polarformer', filename='model_bs_polarformer_float16.ckpt'),
modelConfiguration=HuggingFaceFile(repo_id='hunterhogan/bs_polarformer', filename='bs_polarformer.py'),
modelLogic=hunterFormsBS.BandSplitRotator,
),
Inferred with the class, BandSplitRotator, from:
https://github.com/hunterhogan/hunterFormsBS or https://pypi.org/project/hunterFormsBS/
stereo sequences --111111
[0, 1, 2, 4, 7, 9, 13, 23]
STFT windowing function: Z0Z_tools.halfsineTensor https://github.com/hunterhogan/Z0Z_tools or https://pypi.org/project/Z0Z-tools/
(In other entries, "halfwave" was my codename for the halfsine windowing function.)
N8k3
x @=power
Hunter Hogan
Metrics:
Metric sdr for instrum: 10.0550
Metric si_sdr for instrum: 9.7668
Metric l1_freq for instrum: 39.4597
Metric log_wmse for instrum: 12.2652
Metric aura_stft for instrum: 11.9606
Metric aura_mrstft for instrum: 13.9457
Metric bleedless for instrum: 29.2082
Metric fullness for instrum: 32.0031
Metric sdr for vocals: 10.3512
Metric si_sdr for vocals: 9.8206
Metric l1_freq for vocals: 37.3683
Metric log_wmse for vocals: 12.2652
Metric aura_stft for vocals: 7.0300
Metric aura_mrstft for vocals: 7.9531
Metric bleedless for vocals: 20.5880
Metric fullness for vocals: 20.7692
Date added: 2026-05-21 |
