biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[V1] V1 Enablement Oracle (#13726)

Browse Source 
Signed-off-by: rshaw@neuralmagic.com <rshaw@neuralmagic.com>
Co-authored-by: rshaw@neuralmagic.com <rshaw@neuralmagic.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Michael Goin <michael@neuralmagic.com>
This commit is contained in:
Robert Shaw
2025-03-15 01:02:20 -04:00
committed by GitHub
parent 8c0d15d5c5
commit d4d93db2c5
96 changed files with 1537 additions and 512 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
vllm/v1/engine/llm_engine.py
View File

@@ -46,6 +46,13 @@ class LLMEngine:
use_cached_outputs: bool = False,
multiprocess_mode: bool = False,
) -> None:
if not envs.VLLM_USE_V1:
raise ValueError(
"Using V1 LLMEngine, but envs.VLLM_USE_V1=False. "
"This should not happen. As a workaround, try using "
"LLMEngine.from_vllm_config(...) or explicitly set "
"VLLM_USE_V1=0 or 1 and report this issue on Github.")
self.vllm_config = vllm_config
self.model_config = vllm_config.model_config
self.cache_config = vllm_config.cache_config
@@ -88,6 +95,26 @@ class LLMEngine:
# for v0 compatibility
self.model_executor = self.engine_core.engine_core.model_executor # type: ignore
@classmethod
def from_vllm_config(
cls,
vllm_config: VllmConfig,
usage_context: UsageContext = UsageContext.ENGINE_CONTEXT,
stat_loggers: Optional[dict[str, StatLoggerBase]] = None,
disable_log_stats: bool = False,
) -> "LLMEngine":
if stat_loggers is not None:
raise NotImplementedError(
"Passing StatLoggers to V1 is not yet supported. "
"Set VLLM_USE_V1=0 and file and issue on Github.")
return cls(vllm_config=vllm_config,
executor_class=Executor.get_class(vllm_config),
log_stats=(not disable_log_stats),
usage_context=usage_context,
stat_loggers=stat_loggers,
multiprocess_mode=envs.VLLM_ENABLE_V1_MULTIPROCESSING)
@classmethod
def from_engine_args(
cls,