add default setting for expert parallel size

internlm/core/context/parallel_context.py

@@ -143,7 +143,7 @@ class ParallelContext(metaclass=SingletonMeta):
         self.pipeline_parallel_size = 1
         self.tensor_parallel_size = 1
         self.zero1_parallel_size = -1
-        self.expert_parallel_size = -1
+        self.expert_parallel_size = 1
         self.num_processes_on_current_node = -1
         self.virtual_pipeline_parallel_size = None
         self.virtual_pipeline_parallel_rank = None

internlm/initialize/launch.py

@@ -60,6 +60,9 @@ def args_sanity_check():
     if "tensor" not in gpc.config.parallel:
         gpc.config.parallel._add_item("tensor", 1)
 
+    if "expert" not in gpc.config.parallel:
+        gpc.config.parallel._add_item("expert", 1)
+
     # processing the data config in gpc
     data = gpc.config.data