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

[Misc] Fix Current vLLM config is not set. warnings, assert to avoid issues in the future (#31747)

Browse Source 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
This commit is contained in:
Lucas Wilkinson
2026-01-08 18:20:49 -05:00
committed by GitHub
parent 5d3b6097ad
commit 6cdf015c3c
48 changed files with 380 additions and 240 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/distributed/device_communicators/base_device_communicator.py
View File

@@ -117,9 +117,9 @@ class DeviceCommunicatorBase:
use_ep = False
all2all_backend = None
from vllm.config import get_current_vllm_config
from vllm.config import get_current_vllm_config_or_none
config = get_current_vllm_config()
config = get_current_vllm_config_or_none()
if config is not None:
# as long as we use data parallel (coupled data parallel
# where all data parallel ranks execute forward together),

4
vllm/distributed/device_communicators/quick_all_reduce.py
View File

@@ -9,7 +9,7 @@ from torch.distributed import ProcessGroup
import vllm.envs as envs
from vllm import _custom_ops as ops
from vllm.config import get_current_vllm_config
from vllm.config import get_current_vllm_config_or_none
from vllm.distributed.parallel_state import in_the_same_node_as
from vllm.logger import init_logger
from vllm.platforms import current_platform
@@ -184,7 +184,7 @@ class QuickAllReduce:
)
return
self.qr_quant_level = QuickReduceRegime[regime_str]
vllm_config = get_current_vllm_config()
vllm_config = get_current_vllm_config_or_none()
if (
vllm_config is not None
and hasattr(vllm_config, "model_config")

12
vllm/distributed/parallel_state.py
View File

@@ -1177,9 +1177,9 @@ def init_distributed_environment(
distributed_init_method,
backend,
)
from vllm.config import get_current_vllm_config
from vllm.config import get_current_vllm_config_or_none
config = get_current_vllm_config()
config = get_current_vllm_config_or_none()
if (
config is not None
and config.parallel_config.distributed_executor_backend != "external_launcher"
@@ -1251,7 +1251,7 @@ def init_distributed_environment(
if _WORLD is None:
ranks = list(range(torch.distributed.get_world_size()))
_WORLD = init_world_group(ranks, local_rank, backend)
if config.parallel_config.nnodes > 1:
if config is not None and config.parallel_config.nnodes > 1:
_NODE_COUNT = config.parallel_config.nnodes
else:
_NODE_COUNT = _node_count(_WORLD.cpu_group)
@@ -1260,7 +1260,7 @@ def init_distributed_environment(
assert _WORLD.world_size == torch.distributed.get_world_size(), (
"world group already initialized with a different world size"
)
if config.parallel_config.nnodes_within_dp > 1:
if config is not None and config.parallel_config.nnodes_within_dp > 1:
if parallel_config.data_parallel_size > 1:
world_size_inner_dp = parallel_config.world_size
group_ranks = [
@@ -1316,9 +1316,9 @@ def initialize_model_parallel(
backend = backend or torch.distributed.get_backend(get_world_group().device_group)
data_parallel_size = 1
from vllm.config import get_current_vllm_config
from vllm.config import get_current_vllm_config_or_none
config = get_current_vllm_config()
config = get_current_vllm_config_or_none()
if config is not None:
data_parallel_size = config.parallel_config.data_parallel_size