fix dllm

python/sglang/srt/mem_cache/memory_pool.py

@@ -145,9 +145,10 @@ def available_size(self):
 
     def alloc(self, reqs: list[Req]) -> Optional[List[int]]:
         chunked = [i for i, r in enumerate(reqs) if r.req_pool_idx is not None]
-        assert (
-            len(chunked) <= 1
-        ), "only one chunked request may reuse req_pool_idx in a batch"
+        if not any(r.is_dllm() for r in reqs):
+            assert (
+                len(chunked) <= 1
+            ), "only one chunked request may reuse req_pool_idx in a batch"
         assert all(
             reqs[i].is_chunked > 0 or reqs[i].kv_committed_len > 0 for i in chunked
         ), "request has req_pool_idx but is not chunked"