来自cslt Wiki
(差异) ←上一版本 | 最后版本 (差异) | 下一版本→ (差异)
- sinovoice 11000h "chain" embedded model already outperform current best model !
- prepare the data size / model size experiment for Interspeech.
- plan to start 11000h 7*2048 "chain" model during the holiday.
