(
prompt_kv: int,
reused_kv: int,
generation_kv: int,
)
| 12634 | |
| 12635 | @staticmethod |
| 12636 | def attention_kv_required( |
| 12637 | prompt_kv: int, |
| 12638 | reused_kv: int, |
| 12639 | generation_kv: int, |
| 12640 | ) -> int: |
| 12641 | return prompt_kv - reused_kv + generation_kv |
| 12642 | |
| 12643 | def try_set_sequence_cache_match( |
| 12644 | self, |