Add max_prefill_num_token into server arguments (#133)

This commit is contained in:
Ying Sheng
2024-02-03 02:35:54 -08:00
committed by GitHub
parent 67be11c790
commit e095b16236
3 changed files with 12 additions and 2 deletions

View File

@@ -430,7 +430,8 @@ class Runtime:
load_format: str = "auto",
tokenizer_mode: str = "auto",
trust_remote_code: bool = True,
mem_fraction_static: float = 0.9,
mem_fraction_static: float = ServerArgs.mem_fraction_static,
max_prefill_num_token: int = ServerArgs.max_prefill_num_token,
tp_size: int = 1,
model_mode: List[str] = (),
schedule_heuristic: str = "lpm",
@@ -451,6 +452,7 @@ class Runtime:
tokenizer_mode=tokenizer_mode,
trust_remote_code=trust_remote_code,
mem_fraction_static=mem_fraction_static,
max_prefill_num_token=max_prefill_num_token,
tp_size=tp_size,
model_mode=model_mode,
schedule_heuristic=schedule_heuristic,