A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 12, 09:45 UTC
[Doc] Improve abnormal information string (#29655)
Signed-off-by: maang <maang_h@163.com>
maang-h committed
Nov 28, 2025 at 16:12 UTC
cc0f2a0e19881c3c601d3e287f297b36d2a78f78
1 file changed
+5
-9
vllm/v1/engine/utils.py
+5
-9
@@ -371,8 +371,7 @@ class CoreEngineActorManager:
371
)
372
assert len(nodes) > 0, "No nodes with resources found in Ray cluster."
373
assert dp_master_ip_key in nodes[0], (
374
- "The DP master node (ip: %s) is missing or dead",
375
- dp_master_ip,
374
+ f"The DP master node (ip: {dp_master_ip}) is missing or dead"
375
)
376
device_str = current_platform.ray_device_key
377
n_node_devices: list[int] = [
@@ -446,8 +445,7 @@ class CoreEngineActorManager:
445
if key != "node:__internal_head__" and key.startswith("node:")
446
]
447
assert len(node_ip_keys) == 1, (
449
- "Zero or multiple node IP keys found in node resources: %s",
450
- node_ip_keys,
448
+ f"Zero or multiple node IP keys found in node resources: {node_ip_keys}"
449
)
450
node_ip_key = node_ip_keys[0]
451
node_ip = node_ip_key.split(":")[1]
@@ -464,11 +462,9 @@ class CoreEngineActorManager:
462
if node_ip == dp_master_ip:
463
if dp_size_available < dp_size_local:
464
raise ValueError(
467
- "Not enough resources to allocate %s DP ranks "
468
- "on DP master node %s, possible to fit %s DP ranks",
469
- dp_size_local,
470
- dp_master_ip,
471
- dp_size_available,
465
+ f"Not enough resources to allocate {dp_size_local} DP ranks "
466
+ f"on DP master node {dp_master_ip}, possible to fit "
467
+ f"{dp_size_available} DP ranks."
468
)
469
dp_size_to_allocate = dp_size_local
470
elif pack_strategy == "strict":