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

[Core] Fix engine-use-ray broken (#4105)

Browse Source
This commit is contained in:
SangBin Cho
2024-04-16 14:24:53 +09:00
committed by GitHub
parent 37e84a403d
commit 4e7ee664e2
2 changed files with 16 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
vllm/engine/async_llm_engine.py
View File

@@ -333,8 +333,7 @@ class AsyncLLMEngine:
if engine_config.device_config.device_type == "neuron":
raise NotImplementedError("Neuron is not supported for "
"async engine yet.")
elif (engine_config.parallel_config.worker_use_ray
or engine_args.engine_use_ray):
elif engine_config.parallel_config.worker_use_ray:
initialize_ray_cluster(engine_config.parallel_config)
from vllm.executor.ray_gpu_executor import RayGPUExecutorAsync
executor_class = RayGPUExecutorAsync
@@ -410,8 +409,8 @@ class AsyncLLMEngine:
else:
# FIXME(woosuk): This is a bit hacky. Be careful when changing the
# order of the arguments.
cache_config = args[1]
parallel_config = args[2]
cache_config = kwargs["cache_config"]
parallel_config = kwargs["parallel_config"]
if parallel_config.tensor_parallel_size == 1:
num_gpus = cache_config.gpu_memory_utilization
else: