iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[UX] Improve data-parallel launch validation (#49124)

Browse Source 
Signed-off-by: Taneem Ibrahim <taneem.ibrahim@gmail.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
This commit is contained in:
Taneem Ibrahim
2026-07-24 07:16:39 -07:00
committed by GitHub
co-authored by Wentao Ye
parent 7b40fb9645
commit 453f01783d
1 changed files with 52 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/engine/arg_utils.py
+52 -23
View File
@@ -1955,12 +1955,21 @@ class EngineArgs:
assert not headless or not self.data_parallel_hybrid_lb, (
"data_parallel_hybrid_lb is not applicable in headless mode"
)
assert not (self.data_parallel_hybrid_lb and self.data_parallel_external_lb), (
"data_parallel_hybrid_lb and data_parallel_external_lb cannot both be True."
)
assert self.data_parallel_backend == "mp" or self.nnodes == 1, (
"nnodes > 1 is only supported with data_parallel_backend=mp"
)
if self.data_parallel_hybrid_lb and self.data_parallel_external_lb:
raise ValueError(
"Invalid data-parallel launch options: "
"`--data-parallel-hybrid-lb` and "
"`--data-parallel-external-lb` cannot be enabled together. "
"Enable only one load-balancing mode."
)
if self.nnodes > 1 and self.data_parallel_backend != "mp":
raise ValueError(
"Invalid data-parallel launch options: "
f"`--nnodes {self.nnodes}` requires "
"`--data-parallel-backend mp`; got "
f"`--data-parallel-backend {self.data_parallel_backend}`. "
"Use the MP backend or set `--nnodes 1`."
)
inferred_data_parallel_rank = 0
if self.nnodes > 1:
world_size = (
@@ -1971,13 +1980,22 @@ class EngineArgs:
world_size_within_dp = (
self.pipeline_parallel_size * self.tensor_parallel_size
)
if world_size % self.nnodes != 0:
raise ValueError(
"Invalid data-parallel launch options: "
f"`--nnodes {self.nnodes}` must evenly divide the total "
f"world size ({world_size}). Adjust `--nnodes`, "
"`--data-parallel-size`, `--pipeline-parallel-size`, or "
"`--tensor-parallel-size`."
)
if not 0 <= self.node_rank < self.nnodes:
raise ValueError(
"Invalid data-parallel launch options: `--node-rank` must "
f"be between 0 and {self.nnodes - 1}; got "
f"`--node-rank {self.node_rank}`. Set it to this node's "
"zero-based index."
)
local_world_size = world_size // self.nnodes
assert world_size % self.nnodes == 0, (
f"world_size={world_size} must be divisible by nnodes={self.nnodes}."
)
assert self.node_rank < self.nnodes, (
f"node_rank={self.node_rank} must be less than nnodes={self.nnodes}."
)
inferred_data_parallel_rank = (
self.node_rank * local_world_size
) // world_size_within_dp
@@ -2008,14 +2026,21 @@ class EngineArgs:
)
# Local DP rank = 1, use pure-external LB.
if data_parallel_external_lb:
assert self.data_parallel_rank is not None, (
"data_parallel_rank or node_rank must be specified if "
"data_parallel_external_lb is enable."
)
assert self.data_parallel_size_local in (1, None), (
"data_parallel_size_local must be 1 or None when data_parallel_rank "
"is set"
)
if self.data_parallel_rank is None:
raise ValueError(
"Invalid data-parallel launch options: "
"`--data-parallel-external-lb` requires a data-parallel "
"rank. Set `--data-parallel-rank`, or set "
"`--data-parallel-size` greater than 1 and use `--nnodes` "
"with `--node-rank` so the rank can be inferred."
)
if self.data_parallel_size_local not in (1, None):
raise ValueError(
"Invalid data-parallel launch options: an external "
"data-parallel rank requires `--data-parallel-size-local "
f"1`; got {self.data_parallel_size_local}. Set it to 1 or "
"omit it."
)
data_parallel_size_local = 1
# Use full external lb if we have local_size of 1.
self.data_parallel_hybrid_lb = False
@@ -2050,9 +2075,13 @@ class EngineArgs:
self.node_rank,
)
else:
assert not self.data_parallel_hybrid_lb, (
"data_parallel_size_local must be set to use data_parallel_hybrid_lb."
)
if self.data_parallel_hybrid_lb:
raise ValueError(
"Invalid data-parallel launch options: "
"`--data-parallel-hybrid-lb` requires "
"`--data-parallel-size-local`. Set it to the number of "
"data-parallel ranks on this node."
)
if self.data_parallel_backend == "ray" and (
envs.VLLM_RAY_DP_PACK_STRATEGY == "span"