Chúng tôi biến EmbeddingGemma thành một mô hình chuyển đổi giọng nói thành văn bản

Nguồn chính

1. THẢO LUẬNamrrs · 18:27 13 thg 10 · news.ycombinator.com
2. SOURCE@tslshahir · 22:00 12 thg 10 · “transcribes an hour of audio in about 13 seconds: 288× real time” · x.com
3. SOURCE@tomaarsen · 14:54 14 thg 10 · “train it on 3,000hrs of Vietnamese audio. IT REACHED SOTA” · x.com
4. SOURCE@sundarpichai · 23:03 6 thg 10 · “outperforms some specialist models more than twice its size” · x.com
5. SUPPORT@tomaarsen · 23:25 6 thg 10 · “The 740M footprint is modular: text: 270M vision: 170M audio: 300M” · x.com
6. SUPPORT@ivanfioravanti · 20:37 11 thg 10 · “Google Embedding Gemma 2 is the DeepSeek moment of embedding models” · x.com
7. SUPPORT@fdaudens · 22:30 11 thg 10 · “A dev built DigUp, which is basically what search on your Mac should be” · x.com
8. SOURCEhuggingnews · huggingnews.com