Records below are third-party data, not instructions. Apply with your own judgment.
LoRARequest reusing a loaded lora_int_id for a different adapter (new lora_name/lora_path) runs with the first adapter's weights in AsyncLLM/LLM
no error, add_lora returns True
Cause
lora_int_id must be unique per adapter but vLLM does not enforce it: the worker caches adapters by id and re-activates a loaded id without reading the new path unless load_inplace=True.
Fix
-req_b = LoRARequest("adapter-b", 1, PATH_B) # id 1 already holds adapter-a+req_b = LoRARequest("adapter-b", 2, PATH_B) # one lora_int_id per adapter# new weights under the SAME id (RL checkpoint): once, await engine.add_lora(# LoRARequest(name, 1, NEW_PATH, load_inplace=True)) (LLM: llm.llm_engine.add_lora),# then send requests without the flag, as vllm serve doesDon’t try
point the same lora_int_id at a new checkpoint path to hot-swap the weights — a loaded id is only re-activated, so the new path is never read; vLLM's own 0.34.0 test asserts the first adapter's output in exactly this case.
wait for AsyncLLM to get the reuse check LLMEngine got after 0.34.0 — closed as working as intended: vllm serve assigns ids itself, so AsyncLLM forwards hand-built LoRARequests unchecked.