Support n in OpenAI API completions (#3446)

Co-authored-by: Shan Yu <shanyu1@g.ucla.edu>
Co-authored-by: Yineng Zhang <me@zhyncs.com>
Co-authored-by: chuyue sun <chuyue@lmsys.us-northcentral1-a.compute.internal>

python/sglang/lang/interpreter.py

@@ -566,13 +566,13 @@ class StreamExecutor:
     def _execute_gen(self, expr: SglGen):
         sampling_params = self._resolve_sampling_params(expr.sampling_params)
         name = expr.name
-
         if not self.stream:
             if self.num_api_spec_tokens is None:
                 comp, meta_info = self.backend.generate(
                     self,
                     sampling_params=sampling_params,
                 )
+
             else:
                 if self.backend.is_chat_model:
                     # Speculative execution on models with only chat interface.
@@ -587,8 +587,11 @@ class StreamExecutor:
 
                 else:  # Speculative execution on models with completion interface
                     comp, meta_info = self._spec_gen(sampling_params)
-
-            self.text_ += comp
+            if isinstance(comp, list):
+                self.text_ += comp[0]
+            else:
+                assert isinstance(comp, str)
+                self.text_ += comp
 
             self.variables[name] = comp
             self.meta_info[name] = meta_info
@@ -747,6 +750,7 @@ class StreamExecutor:
         for item in [
             "max_new_tokens",
             "min_new_tokens",
+            "n",
             "stop",
             "stop_token_ids",
             "temperature",