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, 22:15 UTC
[misc] Show driver IP info when Ray fails to allocate driver worker (#13858)
Signed-off-by: Rui Qiao <ruisearch42@gmail.com>
Rui Qiao committed
Feb 25, 2025 at 17:53 UTC
9ba28043b5de4b3b88c6136cdb3b537f56797b64
1 file changed
+4
-3
vllm/executor/ray_distributed_executor.py
+4
-3
@@ -229,9 +229,10 @@ class RayDistributedExecutor(DistributedExecutorBase):
229
logger.debug("driver_dummy_worker: %s", self.driver_dummy_worker)
230
if not self.use_ray_spmd_worker and self.driver_dummy_worker is None:
231
raise ValueError(
232
- "Ray does not allocate any GPUs on the driver node. Consider "
233
- "adjusting the Ray placement group or running the driver on a "
234
- "GPU node.")
232
+ "Ray does not allocate any GPUs on the driver node."
233
+ f"Driver IP: {driver_ip}, worker IPs: {worker_ips}."
234
+ "Consider adjusting the Ray placement group or running "
235
+ "the driver on a GPU node.")
236
237
ip_counts: Dict[str, int] = {}
238
for ip in worker_ips: