Nemotron 3 Diarization โ€” CoreML (Lorqa mirror)

Mirror of FluidInference/nemotron-3-diarization-coreml at revision 25a90f97f254428d4b30374b76af9c74fdee8327, used by the Lorqa app for on-device speaker diarization via FluidAudio (>= 0.17.4). Files are unchanged; the upstream layout is kept.

Path Preset Size
monolithic/v2/Nemotron3Diarizer_fast128.mlmodelc Nemotron3Config.fast128 (10.56 s latency) ~199 MB
split/Nemotron3Diarizer_c128_split_w8a8.mlmodelc c128-split-w8a8 (split graph, W8A8) ~99 MB
learnable_sil_emb.bin silence embedding, required by every preset 2 KB
pre_encode_proj_t.bin pre-encode projection, required by split presets 2 MB

License and attribution

The model is NVIDIA Nemotron 3 Diarization, released under the OpenMDW License Agreement, version 1.1. CoreML conversion by FluidInference.

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. ๐Ÿ™‹ Ask for provider support