GGUF conversion of NVIDIA's
diar_streaming_sortformer_4spk-v2.1
(checkpoint commit fafaab5) for the
transcribe.cpp engine,
produced with the repo's own scripts/convert-sortformer.py at commit 553f109.
Provenance & caveats
Independent conversion for integration testing (Isaree on-device scribe), not
an official artifact of either upstream project.
Q8_0 reproduces the F32 reference speaker boundaries exactly on the engine's
2-speaker fixture; formal DER has not been re-measured on this artifact.
No K-quants by upstream policy: k-tier weight error can deterministically
permute speaker labels mid-stream. Use Q8_0 (or F16).
