misc: fix the req_to_token member change (#967)

python/sglang/srt/managers/tp_worker.py

@@ -289,10 +289,10 @@ class ModelTpServer:
                 "KV cache pool leak detected!"
             )
 
-        if self.req_to_token_pool.can_use_mem_size != self.req_to_token_pool.size:
+        if len(self.req_to_token_pool.free_slots) != self.req_to_token_pool.size:
             warnings.warn(
                 "Warning: "
-                f"available req slots={self.req_to_token_pool.can_use_mem_size}, "
+                f"available req slots={len(self.req_to_token_pool.free_slots)}, "
                 f"total slots={self.req_to_token_pool.size}\n"
                 "Memory pool leak detected!"
             )

python/sglang/srt/mem_cache/memory_pool.py

@@ -32,7 +32,6 @@ class ReqToTokenPool:
         self.req_to_token = torch.empty(
             (size, max_context_len), dtype=torch.int32, device="cuda"
         )
-        self.can_use_mem_size = size
 
     def alloc(self, need_size: int) -> List[int]:
         if need_size > len(self.free_slots):