psymon/gilbeot-korean-audio-litertlm

Update README.md

Update README.md

Correct base CER: 13.14% (device, INT4) — 44% was fabricated, no measurement supports it

Correct dataset attribution: AI-Hub 명령어 음성(노인남녀), 50h/45,823 utt/138 speakers

Add Korean audio LoRA grafted .litertlm (audio_encoder + LM, INT4 same-grid requant)

Add LICENSE with Gemma Terms NOTICE per Section 3.1(4)

Add README with Gemma terms, modifications notice, usage

initial commit