[Bugfix] Register VLLM_BATCH_INVARIANT in envs.py to fix spurious unknown env var warning (#35007)
Signed-off-by: Ranran <1012869439@qq.com> Signed-off-by: Ranran <hzz5361@psu.edu> Signed-off-by: ran <hzz5361@psu.edu> Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
This commit is contained in:
@@ -19,9 +19,6 @@ import torch
|
||||
|
||||
import vllm.envs as envs
|
||||
from vllm.logger import init_logger
|
||||
from vllm.model_executor.layers.batch_invariant import (
|
||||
vllm_is_batch_invariant,
|
||||
)
|
||||
from vllm.platforms import current_platform
|
||||
|
||||
logger = init_logger(__name__)
|
||||
@@ -289,7 +286,7 @@ def supports_trtllm_attention() -> bool:
|
||||
NVIDIA artifactory is accessible, and batch-invariant mode is not enabled.
|
||||
"""
|
||||
# Batch-invariant mode disables TRTLLM attention
|
||||
if vllm_is_batch_invariant():
|
||||
if envs.VLLM_BATCH_INVARIANT:
|
||||
return False
|
||||
|
||||
# Requires SM100 and NVIDIA artifactory to be accessible to download cubins
|
||||
|
||||
Reference in New Issue
Block a user