Instructions to use ALLUCY-Rodent/whisper-small-ko with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ALLUCY-Rodent/whisper-small-ko with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="ALLUCY-Rodent/whisper-small-ko")# Load model directly from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq processor = AutoProcessor.from_pretrained("ALLUCY-Rodent/whisper-small-ko") model = AutoModelForSpeechSeq2Seq.from_pretrained("ALLUCY-Rodent/whisper-small-ko", device_map="auto") - Notebooks
- Google Colab
- Kaggle
whisper-small-ko (q8 ONNX) โ seastar105/whisper-small-komixv2
Meeting AI PWA์ฉ ํ๊ตญ์ด ์์ฑ์ธ์(STT) ๋ชจ๋ธ.
- ์๋ณธ:
seastar105/whisper-small-komixv2(2025-03) โ OpenAI whisper-small์ AI Hub ๋ค์ค ๋๋ฉ์ธ (ํ๊ตญ์ด ์์ฑยท์ฃผ์ยทํ์ ์์ฑยท์ ์์ง ์ ํ๋งยท๋ฐฉ์ก ๋ํ์ฒด)๋ก ํ์ธํ๋. ํ๊ตญ์ด ๋ฒค์น๋งํฌ ํ๊ท WER 7.36 (whisper-large-v3 7.99๋ณด๋ค ์ฐ์, ๋์ผ ํ ์คํธ์ ๊ธฐ์ค). - ๋ณํ: optimum 1.27 + onnxruntime 1.29๋ก int8(q8) ์ง์ ์ต์คํฌํธ + ์์ํ (encoder 93MB / merged decoder 377MB). ort 1.25+ QDQ ํ๊ท ์์ โ ๋ธ๋ผ์ฐ์ ์ธ์ ์์ฑยท์ถ๋ก ๊ฒ์ฆ ์๋ฃ.
- ์ฉ๋: ๋ธ๋ผ์ฐ์ ์จ๋๋ฐ์ด์ค ์ ์ฌ (transformers.js v4, WebGPU/WASM, language=ko)
ํ์ผ ๊ตฌ์กฐ
onnx/
โโโ encoder_model_quantized.onnx # 93MB (q8)
โโโ decoder_model_merged_quantized.onnx # 377MB (q8)
config.json
preprocessor_config.json
tokenizer.json / tokenizer_config.json / added_tokens.json
generation_config.json / special_tokens_map.json / merges.txt / normalizer.json / vocab.json
์ฌ์ฉ (meeting-ai-pwa)
# config.ts ๊ธฐ๋ณธ๊ฐ: ALLUCY-Rodent/whisper-small-ko
VITE_STT_MODEL=ALLUCY-Rodent/whisper-small-ko npm run dev
transformers.js automatic-speech-recognition ํ์ดํ๋ผ์ธ, language: 'ko', dtype q8.
- Downloads last month
- 18