[MISC] Add lora requests to metrics (#9477)

Co-authored-by: Kunjan Patel <kunjanp_google_com@vllm.us-central1-a.c.kunjanp-gke-dev-2.internal>

vllm/engine/llm_engine.py

@@ -1,4 +1,5 @@
 import time
+from collections import Counter as collectionsCounter
 from collections import deque
 from contextlib import contextmanager
 from dataclasses import dataclass
@@ -1617,6 +1618,25 @@ class LLMEngine:
         n_requests: List[int] = []
         finished_reason_requests: List[str] = []
 
+        # Lora requests
+        running_lora_adapters = dict(
+            collectionsCounter([
+                running_request.lora_request.lora_name
+                for scheduler in self.scheduler
+                for running_request in scheduler.running
+                if running_request.lora_request
+            ]))
+        waiting_lora_adapters = dict(
+            collectionsCounter([
+                waiting_request.lora_request.lora_name
+                for scheduler in self.scheduler
+                for waiting_request in scheduler.waiting
+                if waiting_request.lora_request
+            ]))
+        max_lora_stat = "0"
+        if self.lora_config:
+            max_lora_stat = str(self.lora_config.max_loras)
+
         # NOTE: This loop assumes prefill seq_groups are before
         # decode seq_groups in scheduled_seq_groups.
         if scheduler_outputs is not None:
@@ -1738,7 +1758,9 @@ class LLMEngine:
             num_generation_tokens_requests=num_generation_tokens_requests,
             n_requests=n_requests,
             finished_reason_requests=finished_reason_requests,
-        )
+            max_lora=str(max_lora_stat),
+            waiting_lora_adapters=list(waiting_lora_adapters.keys()),
+            running_lora_adapters=list(running_lora_adapters.keys()))
 
     def add_lora(self, lora_request: LoRARequest) -> bool:
         return self.model_executor.add_lora(lora_request)