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, 13:30 UTC
[Misc] Streamline unique id generation (#29375)
Signed-off-by: Nick Hill <nhill@redhat.com>
Nick Hill committed
Nov 25, 2025 at 00:30 UTC
db2906108acdc141e8a21e390228c69b1379e3c2
3 files changed
+16
-13
vllm/entrypoints/openai/protocol.py
+8
-8
@@ -661,7 +661,7 @@ class ChatCompletionRequest(OpenAIBaseModel):
661
),
662
)
663
request_id: str = Field(
664
- default_factory=lambda: f"{random_uuid()}",
664
+ default_factory=random_uuid,
665
description=(
666
"The request_id related to this request. If the caller does "
667
"not set it, a random_uuid will be generated. This id is used "
@@ -1078,7 +1078,7 @@ class CompletionRequest(OpenAIBaseModel):
1078
),
1079
)
1080
request_id: str = Field(
1081
- default_factory=lambda: f"{random_uuid()}",
1081
+ default_factory=random_uuid,
1082
description=(
1083
"The request_id related to this request. If the caller does "
1084
"not set it, a random_uuid will be generated. This id is used "
@@ -1375,7 +1375,7 @@ class EmbeddingCompletionRequest(OpenAIBaseModel):
1375
),
1376
)
1377
request_id: str = Field(
1378
- default_factory=lambda: f"{random_uuid()}",
1378
+ default_factory=random_uuid,
1379
description=(
1380
"The request_id related to this request. If the caller does "
1381
"not set it, a random_uuid will be generated. This id is used "
@@ -1470,7 +1470,7 @@ class EmbeddingChatRequest(OpenAIBaseModel):
1470
),
1471
)
1472
request_id: str = Field(
1473
- default_factory=lambda: f"{random_uuid()}",
1473
+ default_factory=random_uuid,
1474
description=(
1475
"The request_id related to this request. If the caller does "
1476
"not set it, a random_uuid will be generated. This id is used "
@@ -1892,7 +1892,7 @@ class ClassificationCompletionRequest(OpenAIBaseModel):
1892
),
1893
)
1894
request_id: str = Field(
1895
- default_factory=lambda: f"{random_uuid()}",
1895
+ default_factory=random_uuid,
1896
description=(
1897
"The request_id related to this request. If the caller does "
1898
"not set it, a random_uuid will be generated. This id is used "
@@ -1983,7 +1983,7 @@ class ClassificationChatRequest(OpenAIBaseModel):
1983
)
1984
1985
request_id: str = Field(
1986
- default_factory=lambda: f"{random_uuid()}",
1986
+ default_factory=random_uuid,
1987
description=(
1988
"The request_id related to this request. If the caller does "
1989
"not set it, a random_uuid will be generated. This id is used "
@@ -3094,7 +3094,7 @@ class TranslationResponseVerbose(OpenAIBaseModel):
3094
####### Tokens IN <> Tokens OUT #######
3095
class GenerateRequest(BaseModel):
3096
request_id: str = Field(
3097
- default_factory=lambda: f"{random_uuid()}",
3097
+ default_factory=random_uuid,
3098
description=(
3099
"The request_id related to this request. If the caller does "
3100
"not set it, a random_uuid will be generated. This id is used "
@@ -3151,7 +3151,7 @@ class GenerateResponseChoice(BaseModel):
3151
3152
class GenerateResponse(BaseModel):
3153
request_id: str = Field(
3154
- default_factory=lambda: f"{random_uuid()}",
3154
+ default_factory=random_uuid,
3155
description=(
3156
"The request_id related to this request. If the caller does "
3157
"not set it, a random_uuid will be generated. This id is used "
vllm/entrypoints/openai/serving_engine.py
+5
-4
@@ -1349,11 +1349,12 @@ class OpenAIServing:
1349
raw_request: Request | None, default: str | None = None
1350
) -> str | None:
1351
"""Pulls the request id to use from a header, if provided"""
1352
- default = default or random_uuid()
1353
- if raw_request is None:
1354
- return default
1352
+ if raw_request is not None and (
1353
+ (req_id := raw_request.headers.get("X-Request-Id")) is not None
1354
+ ):
1355
+ return req_id
1356
1356
- return raw_request.headers.get("X-Request-Id", default)
1357
+ return random_uuid() if default is None else default
1358
1359
@staticmethod
1360
def _get_data_parallel_rank(raw_request: Request | None) -> int | None:
vllm/utils/__init__.py
+3
-1
@@ -52,9 +52,11 @@ STR_FLASHINFER_ATTN_VAL: str = "FLASHINFER"
52
STR_FLASH_ATTN_VAL: str = "FLASH_ATTN"
53
STR_INVALID_VAL: str = "INVALID"
54
55
+MASK_64_BITS = (1 << 64) - 1
56
+
57
58
def random_uuid() -> str:
57
- return str(uuid.uuid4().hex)
59
+ return f"{uuid.uuid4().int & MASK_64_BITS:016x}" # 16 hex chars
60
61
62
def length_from_prompt_token_ids_or_embeds(