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 15, 08:30 UTC
[Misc] Reduce logs on startup (#18649)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
May 25, 2025 at 14:03 UTC
503f8487c295343ac13124bc2483cee4e2aabf02
3 files changed
+16
-16
vllm/model_executor/layers/quantization/fp8.py
+2
-3
@@ -62,10 +62,9 @@ class Fp8Config(QuantizationConfig):
62
weight_block_size: Optional[list[int]] = None,
63
) -> None:
64
super().__init__()
65
+
66
self.is_checkpoint_fp8_serialized = is_checkpoint_fp8_serialized
66
- if is_checkpoint_fp8_serialized:
67
- logger.warning("Detected fp8 checkpoint. Please note that the "
68
- "format is experimental and subject to change.")
67
+
68
if activation_scheme not in ACTIVATION_SCHEMES:
69
raise ValueError(
70
f"Unsupported activation scheme {activation_scheme}")
vllm/platforms/__init__.py
+1
-4
@@ -217,11 +217,8 @@ def resolve_current_platform_cls_qualname() -> str:
217
platform_cls_qualname = func()
218
if platform_cls_qualname is not None:
219
activated_plugins.append(name)
220
- logger.info("Platform plugin %s loaded.", name)
221
- logger.warning(
222
- "Platform plugin %s function's return value is None", name)
220
except Exception:
224
- logger.exception("Failed to load platform plugin %s", name)
221
+ pass
222
223
activated_builtin_plugins = list(
224
set(activated_plugins) & set(builtin_platform_plugins.keys()))
vllm/plugins/__init__.py
+13
-9
@@ -2,7 +2,7 @@
2
3
import logging
4
import os
5
-from typing import Callable
5
+from typing import Any, Callable
6
7
import torch
8
@@ -14,7 +14,7 @@ logger = logging.getLogger(__name__)
14
plugins_loaded = False
15
16
17
-def load_plugins_by_group(group: str) -> dict[str, Callable]:
17
+def load_plugins_by_group(group: str) -> dict[str, Callable[[], Any]]:
18
import sys
19
if sys.version_info < (3, 10):
20
from importlib_metadata import entry_points
@@ -27,23 +27,27 @@ def load_plugins_by_group(group: str) -> dict[str, Callable]:
27
if len(discovered_plugins) == 0:
28
logger.debug("No plugins for group %s found.", group)
29
return {}
30
+
31
logger.info("Available plugins for group %s:", group)
32
for plugin in discovered_plugins:
32
- logger.info("name=%s, value=%s", plugin.name, plugin.value)
33
+ logger.info("- %s -> %s", plugin.name, plugin.value)
34
+
35
if allowed_plugins is None:
34
- logger.info("all available plugins for group %s will be loaded.",
35
- group)
36
- logger.info("set environment variable VLLM_PLUGINS to control"
37
- " which plugins to load.")
38
- plugins = {}
36
+ logger.info("All plugins in this group will be loaded. "
37
+ "Set `VLLM_PLUGINS` to control which plugins to load.")
38
+
39
+ plugins = dict[str, Callable[[], Any]]()
40
for plugin in discovered_plugins:
41
if allowed_plugins is None or plugin.name in allowed_plugins:
42
+ if allowed_plugins is not None:
43
+ logger.info("Loading plugin %s", plugin.name)
44
+
45
try:
46
func = plugin.load()
47
plugins[plugin.name] = func
44
- logger.info("plugin %s loaded.", plugin.name)
48
except Exception:
49
logger.exception("Failed to load plugin %s", plugin.name)
50
+
51
return plugins
52
53