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

[misc] soft drop beam search (#8763)

Browse Source
This commit is contained in:
youkaichao 2024-09-24 15:48:39 -07:00 committed by GitHub
parent 2467b642dd
commit 1e7d5c01f5
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 10 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/envs.py
View File

@ -62,6 +62,7 @@ if TYPE_CHECKING:
VLLM_TORCH_PROFILER_DIR: Optional[str] = None VLLM_TORCH_PROFILER_DIR: Optional[str] = None
VLLM_USE_TRITON_AWQ: bool = False VLLM_USE_TRITON_AWQ: bool = False
VLLM_ALLOW_RUNTIME_LORA_UPDATING: bool = False VLLM_ALLOW_RUNTIME_LORA_UPDATING: bool = False
VLLM_ALLOW_DEPRECATED_BEAM_SEARCH: bool = False
def get_default_cache_root(): def get_default_cache_root():
@ -195,6 +196,10 @@ environment_variables: Dict[str, Callable[[], Any]] = {
lambda: (os.environ.get("VLLM_USE_TRITON_FLASH_ATTN", "True").lower() in lambda: (os.environ.get("VLLM_USE_TRITON_FLASH_ATTN", "True").lower() in
("true", "1")), ("true", "1")),
# If set, allowing the use of deprecated beam search implementation
"VLLM_ALLOW_DEPRECATED_BEAM_SEARCH":
lambda: os.environ.get("VLLM_ALLOW_DEPRECATED_BEAM_SEARCH", "0") == "1",
# Internal flag to enable Dynamo graph capture # Internal flag to enable Dynamo graph capture
"VLLM_TEST_DYNAMO_GRAPH_CAPTURE": "VLLM_TEST_DYNAMO_GRAPH_CAPTURE":
lambda: int(os.environ.get("VLLM_TEST_DYNAMO_GRAPH_CAPTURE", "0")), lambda: int(os.environ.get("VLLM_TEST_DYNAMO_GRAPH_CAPTURE", "0")),

5
vllm/sampling_params.py
View File

@ -8,6 +8,7 @@ import msgspec
import torch import torch
from typing_extensions import Annotated from typing_extensions import Annotated
import vllm.envs as envs
from vllm.logger import init_logger from vllm.logger import init_logger
logger = init_logger(__name__) logger = init_logger(__name__)
@ -260,6 +261,10 @@ class SamplingParams(
self._verify_args() self._verify_args()
if self.use_beam_search: if self.use_beam_search:
if not envs.VLLM_ALLOW_DEPRECATED_BEAM_SEARCH:
raise ValueError(
"Using beam search as a sampling parameter is deprecated, and will be removed in the future release. Please use the `vllm.LLM.use_beam_search` method for dedicated beam search instead, or set the environment variable `VLLM_ALLOW_DEPRECATED_BEAM_SEARCH=1` to suppress this error. For more details, see https://github.com/vllm-project/vllm/issues/8306 ." # noqa
)
self._verify_beam_search() self._verify_beam_search()
else: else:
self._verify_non_beam_search() self._verify_non_beam_search()