cleanup: remove pre-split model paths

Docs and run script for the two model variants

Docs and run script for the two model variants

Docs and run script for the two model variants

Add AMI-finetuned quantized model; split models/ by variant

Add AMI-finetuned quantized model; split models/ by variant

Add AMI-finetuned quantized model; split models/ by variant

Add AMI-finetuned quantized model; split models/ by variant

Read speaker channel count from the model instead of assuming 10

Read speaker channel count from the model instead of assuming 10

Read speaker channel count from the model instead of assuming 10

README: both collar conventions, AMI Eval results, length-dependent degradation

bin: rebuild with cache flush/invalidate and band-limited resampling

README: correct C++ metrics, fix GitHub link, document 8 kHz requirement

cleanup: remove local upload helper

Create config.json

LS-EEND streaming speaker diarization for AX650N: axmodel, Python SDK, prebuilt C++ binary

initial commit