A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 23:30 UTC
[V1][PP] Enable true PP with Ray executor (#13472)
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Woosuk Kwon committed
Feb 18, 2025 at 09:15 UTC
4fb8142a0e47c83c5f98fcf76f1168d6f6057ff3
1 file changed
+1
-1
vllm/v1/executor/ray_distributed_executor.py
+1
-1
@@ -32,7 +32,7 @@ class RayDistributedExecutor(RayDistributedExecutorV0, Executor):
32
"""Ray distributed executor supports pipeline parallelism,
33
meaning that it allows PP size batches to be executed concurrently.
34
"""
35
- return 1 #self.vllm_config.parallel_config.pipeline_parallel_size
35
+ return self.parallel_config.pipeline_parallel_size
36
37
def execute_model(
38
self,