Support limiting max loaded loras in CPU. (#8650)

2025-08-03 00:02:23 -07:00
parent a437aa9987
commit 8675bdf246
8 changed files with 163 additions and 58 deletions
--- a/python/sglang/srt/lora/lora_registry.py
+++ b/python/sglang/srt/lora/lora_registry.py
@@ -186,3 +186,10 @@ class LoRARegistry:
        self._registry[lora_ref.lora_name] = lora_ref
        self._counters[lora_ref.lora_id] = ConcurrentCounter()
        return lora_ref
+
+    @property
+    def num_registered_loras(self) -> int:
+        """
+        Returns the total number of LoRA adapters currently registered.
+        """
+        return len(self._registry)