[PD] Change kv_load_failure_policy Default from "recompute" to "fail" (#34896)

Signed-off-by: NickLucche <nlucches@redhat.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
This commit is contained in:
Nicolò Lucchesi
2026-02-21 01:34:57 -08:00
committed by GitHub
co-authored by Cyrus Leung
parent 8dc8a99b56
commit ab6f3487a6
5 changed files with 10 additions and 7 deletions
@@ -30,7 +30,7 @@ def _make_get_num_new_matched_tokens(
@pytest.fixture
def scheduler():
vllm_config = create_vllm_config()
vllm_config = create_vllm_config(kv_load_failure_policy="recompute")
return create_scheduler(vllm_config)
+3 -1
View File
@@ -5,7 +5,7 @@ from collections import defaultdict
from collections.abc import Callable
from dataclasses import dataclass
from itertools import chain, count
from typing import Any
from typing import Any, Literal
import torch
@@ -96,6 +96,7 @@ def create_vllm_config(
cache_dtype: str = "auto",
hf_overrides: dict[str, Any] | None = None,
attention_backend: str | None = None,
kv_load_failure_policy: Literal["recompute", "fail"] = "fail",
) -> VllmConfig:
"""Initialize VllmConfig For Testing."""
model_config = ModelConfig(
@@ -125,6 +126,7 @@ def create_vllm_config(
kv_role="kv_both",
enable_permute_local_kv=enable_permute_local_kv,
kv_connector_extra_config=kv_connector_extra_config or {},
kv_load_failure_policy=kv_load_failure_policy,
)
attention_config = AttentionConfig(backend=attention_backend)
return VllmConfig(