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, 17:15 UTC
[Misc] Remove ignore_reinit_error for ray.init() (#15373)
Rui Qiao committed
Mar 24, 2025 at 00:41 UTC
5797fb97e9df1ca5ef1469a041946e05b896cfb8
1 file changed
+3
-5
vllm/executor/ray_utils.py
+3
-5
@@ -289,16 +289,14 @@ def initialize_ray_cluster(
289
elif current_platform.is_rocm() or current_platform.is_xpu():
290
# Try to connect existing ray instance and create a new one if not found
291
try:
292
- ray.init("auto", ignore_reinit_error=True)
292
+ ray.init("auto")
293
except ConnectionError:
294
logger.warning(
295
"No existing RAY instance detected. "
296
"A new instance will be launched with current node resources.")
297
- ray.init(address=ray_address,
298
- ignore_reinit_error=True,
299
- num_gpus=parallel_config.world_size)
297
+ ray.init(address=ray_address, num_gpus=parallel_config.world_size)
298
else:
301
- ray.init(address=ray_address, ignore_reinit_error=True)
299
+ ray.init(address=ray_address)
300
301
device_str = current_platform.ray_device_key
302
if not device_str: