xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-07 00:27:07 +08:00
Code Issues Packages Projects Releases Wiki Activity

temporarily remove enable_microbatching

Browse Source 
Signed-off-by: Sage Moore <sage@neuralmagic.com>
This commit is contained in:
Sage Moore 2025-06-02 19:28:58 +00:00
parent e34e4411b9
commit 919eef995b
1 changed files with 0 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
vllm/config.py
View File

@ -1741,9 +1741,6 @@ class ParallelConfig:
rank: int = 0
"""Global rank in distributed setup."""
enable_microbatching: bool = False
"""Enable microbatching for the model executor."""
always_microbatch_if_enabled: bool = True
"""Always microbatch if microbatching is enabled. Easier to sync between
dp workers."""
@ -4325,16 +4322,6 @@ class VllmConfig:
"cascade attention. Disabling cascade attention.")
self.model_config.disable_cascade_attn = True
if self.parallel_config.enable_microbatching and \
self.compilation_config.level >= CompilationLevel.PIECEWISE:
# Microbatching is not supported with piecewise compilation yet.
# More specifically piecewise cuda-graphs
logger.warning_once(
"Piecewise compilation is not supported with "
"microbatching. Disabling piecewiseching compilation.")
self.compilation_config.level = CompilationLevel.NO_COMPILATION
if self.model_config and self.model_config.use_mla and \
not (current_platform.is_cuda() or current_platform.is_rocm()):
logger.info(