mirror of
https://github.com/hanzoai/koe.git
synced 2026-10-30 18:33:00 +00:00
Jaeyong SungandGitHub
bd24e1ba40
* refactor: enable gqa, remove dropout * remove unnecessary code * change kv cache * refactor: faster inference speed with lower precision support - larger torch.compile - support fp16, bf16, fp32 inference - refactor model for better code quality * fix merge
7 lines
50 B
Python
7 lines
50 B
Python
from .model import Dia
|
|
|
|
|
|
__all__ = [
|
|
"Dia",
|
|
]
|