ECAPA-TDNN speaker verification — ONNX export

This is an ONNX re-export of speechbrain/spkrec-ecapa-voxceleb. The model is SpeechBrain's; only the export is ours.

It exists so that Astrolabos Voice can perform speaker verification on a fresh machine without a PyTorch install. Every other model the voice daemon needs is already fetchable; this one was not, which meant setup could not complete unaided anywhere but the machine it was built on.

Attribution and licence

Source model: speechbrain/spkrec-ecapa-voxceleb — SpeechBrain
Licence: Apache-2.0, inherited from the source model
Changes: exported to ONNX. No retraining, no fine-tuning, no change to weights.

If you are choosing a model rather than running Astrolabos, use the source model. This export is a packaging convenience, not an improvement on it.

Provenance

Exported with scripts/export-ecapa-onnx.py from the Astrolabos Voice repository.


file	`ecapa_tdnn.onnx`
size	83,404,680 bytes
sha256	`9acb086d5d9efca07266930ce03772f13cad5fd5a415bcf014ee4d82bed26311`

The consuming code verifies this checksum on download; a mismatch is a hard failure rather than a warning.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for astrolabos/ecapa-tdnn

Base model

speechbrain/spkrec-ecapa-voxceleb

Quantized

(4)

this model