From af6e19f50f1d5d0c3801948c3ab17b2af231c259 Mon Sep 17 00:00:00 2001 From: wenxindongwork <161090399+wenxindongwork@users.noreply.github.com> Date: Sat, 1 Nov 2025 11:14:44 -0600 Subject: [PATCH] [Core][TPU] Support TPU Data Parallalism (#27365) Signed-off-by: wenxindongwork --- vllm/entrypoints/llm.py | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/vllm/entrypoints/llm.py b/vllm/entrypoints/llm.py index 758e16c89e694..b0b996ab2fec5 100644 --- a/vllm/entrypoints/llm.py +++ b/vllm/entrypoints/llm.py @@ -67,6 +67,7 @@ from vllm.outputs import ( RequestOutput, ScoringRequestOutput, ) +from vllm.platforms import current_platform from vllm.pooling_params import PoolingParams from vllm.sampling_params import BeamSearchParams, RequestOutputKind, SamplingParams from vllm.tasks import PoolingTask @@ -289,7 +290,11 @@ class LLM: # warn about single-process data parallel usage. _dp_size = int(kwargs.get("data_parallel_size", 1)) _distributed_executor_backend = kwargs.get("distributed_executor_backend") - if _dp_size > 1 and not _distributed_executor_backend == "external_launcher": + if ( + _dp_size > 1 + and not _distributed_executor_backend == "external_launcher" + and not current_platform.is_tpu() + ): raise ValueError( f"LLM(data_parallel_size={_dp_size}) is not supported for single-" "process usage and may hang. Please use "