SepRQ : Self-Supervised Speech Mixture Representation Learning Via Mask-Free Multi-Scale Source Separation
SSL speech feature extractors for SepRQ and BEST-RQ (50 Hz). The
encoder (CNN frontend + 12-layer Conformer) and global norm stats are pulled
from the Hugging Face Hub (SevKod/SepRQ)
on first use — only the requested model is downloaded. Upstream
speechbrain>=1.0.3 only; runs on CPU or GPU.
Install
pip install seprq
# or from the git repo
pip install git+https://github.com/SevKod/SepRQ.git # install straight from GitHub
Use
from seprq import SepRQEncoder
speech_encoder = SepRQEncoder("SepRQ") # or "BestRQ_50Hz"
layers = speech_encoder("utterance.wav") # list of 12 tensors, each [1, T, 576]
final = layers[-1] # last Conformer layer
Calling the encoder runs forward and returns the 12 Conformer layer
outputs as a list, each [1, T, 576].
It accepts a file path (any format/sample rate — decoded, downmixed to mono and resampled to 16 kHz for you) or a raw 1D 16 kHz waveform (numpy array / tensor).
The repo is private, so authenticate once: huggingface-cli login (or set HF_TOKEN).
1 maintainer
Metadata
1 maintainer
Release files for seprq 0.1.3
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| seprq-0.1.3.tar.gz | 3.4 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| seprq-0.1.3-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 7.0 kB
Release files / seprq-0.1.3.tar.gz
| Download URL | seprq-0.1.3.tar.gz |
|---|---|
| Size | 3.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
73a2a9d9ce18a38fb4ab04f1df5a72f77e76d9c002eb6944b926e0c3e2288355
|
|
BLAKE2b-256 checksum How to use checksums |
80a633b5b0f4d5ae1ab109da19a9fda6629f7f5a1351d2282018b453643e22fd
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 30, 2026.
Transparency logRelease files / seprq-0.1.3-py3-none-any.whl
| Download URL | seprq-0.1.3-py3-none-any.whl |
|---|---|
| Size | 3.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
fa60f926988a100106df39cdb2ddf29186caa6d316263c92efe71a4e4b4a5e86
|
|
BLAKE2b-256 checksum How to use checksums |
dc4a438bd3955e36f155a1e60d4b3cfef50f57863f612e2ea00010160e81a061
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 30, 2026.
Transparency log
