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 8, 05:00 UTC
[Bugfix] Fix 'remove_instance_endpoint' method logic in disagg_proxy_demo (#32922)
Signed-off-by: ChenqianCao <39755070+ChenqianCao@users.noreply.github.com> Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
ChenqianCao committed
Feb 17, 2026 at 18:06 UTC
ad65177a1977af535ac0f82312f898a03c425632
1 file changed
+2
-2
examples/online_serving/disaggregated_serving/disagg_proxy_demo.py
+2
-2
@@ -328,9 +328,9 @@ class Proxy:
328
if instance_type == "decode" and instance in self.decode_instances:
329
self.decode_instances.remove(instance)
330
self.decode_cycler = itertools.cycle(self.decode_instances)
331
- if instance_type == "prefill" and instance in self.decode_instances:
331
+ if instance_type == "prefill" and instance in self.prefill_instances:
332
self.prefill_instances.remove(instance)
333
- self.prefill_cycler = itertools.cycle(self.decode_instances)
333
+ self.prefill_cycler = itertools.cycle(self.prefill_instances)
334
335
336
class RoundRobinSchedulingPolicy(SchedulingPolicy):