[Bugfix] Fix KeyError: 1 When Using LoRA adapters (vllm-project#5164)

xjpang · Jul 8, 2024 · dd9ec19 · dd9ec19
1 parent dbc803a
commit dd9ec19
Showing 1 changed file with 3 additions and 1 deletion.
diff --git a/vllm/core/scheduler.py b/vllm/core/scheduler.py
@@ -423,7 +423,9 @@ def _schedule_running(
                 num_running_seqs = seq_group.get_max_num_running_seqs()
                 budget.subtract_num_seqs(seq_group.request_id,
                                          num_running_seqs)
-                if curr_loras is not None and seq_group.lora_int_id > 0:
+
+                if (curr_loras is not None and seq_group.lora_int_id > 0
+                        and seq_group.lora_int_id in curr_loras):
                     curr_loras.remove(seq_group.lora_int_id)
 
                 if running_queue: