.

app/main.py

@@ -217,29 +217,18 @@ def _worker(t, q):
 
 
 def _encode(texts: List[str]):
-    """
+    try:
-    单次请求：
+        return model.encode(texts, return_dense=True)
-      1. 子进程跑 GPU 推理；成功→返回
+    except RuntimeError as e:
-      2. 若子进程 OOM / CUDA Error → 同一次请求 fallback 到 CPU
+        if "out of memory" in str(e).lower() or "cuda error" in str(e).lower():
-    绝不改全局状态，其他并发请求不受影响
+            logger.warning("GPU OOM → fallback to CPU: %s", str(e))
-    """
-    q = mp.Queue()
-    p = mp.Process(target=_worker, args=(texts, q))
-    p.start()
-    p.join(timeout=60)
-
-    if not q.empty():
-        status, payload = q.get()
-        if status == "ok":
-            return payload
-        if "out of memory" in payload.lower() or "cuda error" in payload.lower():
-            logger.warning("GPU OOM → 本次请求改走 CPU：%s", payload)
             torch.cuda.empty_cache()
-            cpu_model, _ = load_model("cpu")
+            global CPU_MODEL_CACHE
-            return cpu_model.encode(texts, return_dense=True)
+            if CPU_MODEL_CACHE is None:
-        raise RuntimeError(payload)
+                CPU_MODEL_CACHE, _ = load_model("cpu")
+            return CPU_MODEL_CACHE.encode(texts, return_dense=True)
+        raise
 
-    raise RuntimeError("子进程异常退出，无返回")
 
 
 @app.post("/v1/embeddings")