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

[Misc] Remove deprecated profiler environment variables (#33536)

Browse Source 
Signed-off-by: carlory <baofa.fan@daocloud.io>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
杨朱 · Kiki
2026-02-03 14:58:44 +08:00
committed by GitHub
parent e10604480b
commit ef248ff740
2 changed files with 0 additions and 95 deletions
Download Patch File Download Diff File Expand all files Collapse all files

81
vllm/config/profiler.py
View File

@@ -8,7 +8,6 @@ from pydantic import Field, model_validator
from pydantic.dataclasses import dataclass
from typing_extensions import Self
import vllm.envs as envs
from vllm.config.utils import config
from vllm.logger import init_logger
from vllm.utils.hashing import safe_hash
@@ -102,88 +101,8 @@ class ProfilerConfig:
hash_str = safe_hash(str(factors).encode(), usedforsecurity=False).hexdigest()
return hash_str
def _get_from_env_if_set(self, field_name: str, env_var_name: str) -> None:
"""Get field from env var if set, with deprecation warning."""
if envs.is_set(env_var_name):
value = getattr(envs, env_var_name)
logger.warning_once(
"Using %s environment variable is deprecated and will be removed in "
"v0.15.0 or v1.0.0, whichever is soonest. Please use "
"--profiler-config.%s command line argument or "
"ProfilerConfig(%s=...) config field instead.",
env_var_name,
field_name,
field_name,
)
return value
return None
def _set_from_env_if_set(
self,
field_name: str,
env_var_name: str,
to_bool: bool = True,
to_int: bool = False,
) -> None:
"""Set field from env var if set, with deprecation warning."""
value = self._get_from_env_if_set(field_name, env_var_name)
if value is not None:
if to_bool:
value = value == "1"
if to_int:
value = int(value)
setattr(self, field_name, value)
@model_validator(mode="after")
def _validate_profiler_config(self) -> Self:
maybe_use_cuda_profiler = self._get_from_env_if_set(
"profiler", "VLLM_TORCH_CUDA_PROFILE"
)
if maybe_use_cuda_profiler is not None:
self.profiler = "cuda" if maybe_use_cuda_profiler == "1" else None
else:
self._set_from_env_if_set(
"torch_profiler_dir", "VLLM_TORCH_PROFILER_DIR", to_bool=False
)
if self.torch_profiler_dir:
self.profiler = "torch"
self._set_from_env_if_set(
"torch_profiler_record_shapes",
"VLLM_TORCH_PROFILER_RECORD_SHAPES",
)
self._set_from_env_if_set(
"torch_profiler_with_memory",
"VLLM_TORCH_PROFILER_WITH_PROFILE_MEMORY",
)
self._set_from_env_if_set(
"torch_profiler_with_stack",
"VLLM_TORCH_PROFILER_WITH_STACK",
)
self._set_from_env_if_set(
"torch_profiler_with_flops",
"VLLM_TORCH_PROFILER_WITH_FLOPS",
)
self._set_from_env_if_set(
"ignore_frontend",
"VLLM_TORCH_PROFILER_DISABLE_ASYNC_LLM",
)
self._set_from_env_if_set(
"torch_profiler_use_gzip",
"VLLM_TORCH_PROFILER_USE_GZIP",
)
self._set_from_env_if_set(
"torch_profiler_dump_cuda_time_total",
"VLLM_TORCH_PROFILER_DUMP_CUDA_TIME_TOTAL",
)
self._set_from_env_if_set(
"delay_iterations", "VLLM_PROFILER_DELAY_ITERS", to_bool=False, to_int=True
)
self._set_from_env_if_set(
"max_iterations", "VLLM_PROFILER_MAX_ITERS", to_bool=False, to_int=True
)
has_delay_or_limit = self.delay_iterations > 0 or self.max_iterations > 0
if self.profiler == "torch" and has_delay_or_limit and not self.ignore_frontend:
logger.warning_once(

14
vllm/envs.py
View File

@@ -88,20 +88,6 @@ if TYPE_CHECKING:
VLLM_PLUGINS: list[str] | None = None
VLLM_LORA_RESOLVER_CACHE_DIR: str | None = None
VLLM_LORA_RESOLVER_HF_REPO_LIST: str | None = None
# Deprecated env variables for profiling, kept for backward compatibility
# See also vllm/config/profiler.py and `--profiler-config` argument
VLLM_TORCH_CUDA_PROFILE: str | None = None
VLLM_TORCH_PROFILER_DIR: str | None = None
VLLM_TORCH_PROFILER_RECORD_SHAPES: str | None = None
VLLM_TORCH_PROFILER_WITH_PROFILE_MEMORY: str | None = None
VLLM_TORCH_PROFILER_DISABLE_ASYNC_LLM: str | None = None
VLLM_TORCH_PROFILER_WITH_STACK: str | None = None
VLLM_TORCH_PROFILER_WITH_FLOPS: str | None = None
VLLM_TORCH_PROFILER_USE_GZIP: str | None = None
VLLM_TORCH_PROFILER_DUMP_CUDA_TIME_TOTAL: str | None = None
VLLM_PROFILER_DELAY_ITERS: str | None = None
VLLM_PROFILER_MAX_ITERS: str | None = None
# End of deprecated env variables for profiling
VLLM_USE_AOT_COMPILE: bool = False
VLLM_USE_BYTECODE_HOOK: bool = False
VLLM_FORCE_AOT_LOAD: bool = False