fix strategy (#46256)

python/paddle/distributed/auto_parallel/constants.py

@@ -45,7 +45,7 @@ set_field_default_config(BASE, "gradient_scale", True)
 set_field_default_config(BASE, "use_cache", True)
 set_field_default_config(BASE, "return_numpy", True)
 set_field_default_config(BASE, "all_ranks", False)
-set_field_default_config(BASE, "split_data", False)
+set_field_default_config(BASE, "split_data", True)
 set_field_default_config(BASE, "seed", None)
 set_field_default_config(BASE, "reinit", False)  # Only for debug
 

python/paddle/distributed/auto_parallel/engine.py

@@ -978,9 +978,10 @@ class Engine:
 
         # extract ckpts by specific model
         if isinstance(self._model, paddle.nn.Layer):
-            if hasattr(
-                    self._model, "gpt"
-            ) and self._model.__class__.__name__ == 'GPTForPretraining':
+            if hasattr(self._model,
+                       "gpt") and self._model.__class__.__name__ in [
+                           'GPTForPretraining', 'GPTForPretrainingAuto'
+                       ]:
                 exact_ckpts = self._model.gpt.checkpoints
             else:
                 exact_ckpts = recompute.checkpoints

python/paddle/distributed/auto_parallel/strategy.py

@@ -59,10 +59,11 @@ class BaseConfig(object):
         return result_dict
 
     def __repr__(self):
-        return yaml.dump(self.to_dict(),
-                         default_flow_style=False,
-                         sort_keys=True,
-                         indent=4)
+        result_dict = self.to_dict()
+        string = "{"
+        for k, v in result_dict.items():
+            string += "\"%s\":\"%s\"," % (k, v)
+        return string + "}"
 
     def __deepcopy__(self, memo):
         cls = self.__class__

python/paddle/distributed/auto_parallel/utils.py

@@ -81,6 +81,8 @@ def convert_to_dims_mapping(shard_spec, process_mesh):
     for shard in shard_spec:
         if shard is None:
             dims_mapping.append(-1)
+        elif process_mesh.topology[process_mesh.dim_names.index(shard)] == 1:
+            dims_mapping.append(-1)
         else:
             dims_mapping.append(process_mesh.dim_names.index(shard))
     return dims_mapping