Merge pull request #16954 from junjun315/fix-dygraph-checkpoint

Fix dygraph checkpoint bug

paddle/fluid/imperative/layer.h

@@ -464,7 +464,11 @@ class PYBIND11_HIDDEN RuntimeInferVarTypeContext
 
   void SetType(const std::string& name,
                framework::proto::VarType::Type type) override {
-    var_set_[name]->SetType(type);
+    if (name == "kLookupTablePath") {
+      VLOG(2) << "SUPER UGLY FIX, remove this when move imperative mode in C++";
+    } else {
+      var_set_[name]->SetType(type);
+    }
   }
 
   framework::proto::VarType::Type GetDataType(

python/paddle/fluid/dygraph/checkpoint.py

@@ -113,14 +113,17 @@ def load_persistables(vardict, dirname, filename=None):
 def _save_var_to_file(stat_dict, file_dir, file_name):
     save_block = default_main_program().global_block()
     save_var_map = {}
-    for each_var in stat_dict.items():
+    for var_key, each_var in stat_dict.items():
         save_var_map[each_var.name] = each_var
         if file_name is None:
             save_block.append_op(
                 type='save',
                 inputs={'X': [each_var]},
                 outputs={},
-                attrs={'file_path': os.path.join(file_dir, each_var.name)})
+                attrs={
+                    'file_path': os.path.join(file_dir,
+                                              os.path.normpath(each_var.name))
+                })
 
     if file_name is not None:
         save_var_list = []
@@ -131,14 +134,16 @@ def _save_var_to_file(stat_dict, file_dir, file_name):
             type='save_combine',
             inputs={'X': save_var_list},
             outputs={},
-            attrs={'file_path': os.path.join(file_dir, file_name)})
+            attrs={
+                'file_path': os.path.join(file_dir, os.path.normpath(file_name))
+            })
 
 
 def _load_var_from_file(stat_dict, file_dir, file_name):
     load_block = default_main_program().global_block()
     load_var_map = {}
 
-    for each_var in stat_dict.items():
+    for var_key, each_var in stat_dict.items():
         assert isinstance(each_var, Variable)
         if each_var.type == core.VarDesc.VarType.RAW:
             continue
@@ -148,7 +153,10 @@ def _load_var_from_file(stat_dict, file_dir, file_name):
                 type='load',
                 inputs={},
                 outputs={'Out': [new_var]},
-                attrs={'file_path': os.path.join(file_dir, each_var.name)})
+                attrs={
+                    'file_path': os.path.join(file_dir,
+                                              os.path.normpath(each_var.name))
+                })
 
         load_var_map[new_var.name] = new_var
 
@@ -161,7 +169,9 @@ def _load_var_from_file(stat_dict, file_dir, file_name):
             type='load_combine',
             inputs={},
             outputs={"Out": load_var_list},
-            attrs={'file_path': os.path.join(file_dir, file_name)})
+            attrs={
+                'file_path': os.path.join(file_dir, os.path.normpath(file_name))
+            })
         for res_var in load_var_list:
             load_var_map[res_var.name] = res_var
 
@@ -175,5 +185,5 @@ def _clone_var_in_block_(block, var):
         shape=var.shape,
         dtype=var.dtype,
         type=var.type,
-        lod_level=var.lod_level,
+        lod_level=0,
         persistable=True)

python/paddle/fluid/dygraph/layers.py

@@ -246,7 +246,10 @@ class Layer(core.Layer):
     def load_dict(self, stat_dict, include_sublayers=True):
         for name, item in self.__dict__.get('_parameters', None).items():
             if item.name in stat_dict:
-                self.__setattr__(name, stat_dict[item.name])
+                var = item._ivar.value()
+                tensor = var.get_tensor()
+                tensor.set(stat_dict[item.name].numpy(),
+                           framework._current_expected_place())
 
         if include_sublayers:
             for layer_name, layer_item in self._sub_layers.items():

python/paddle/fluid/tests/unittests/test_imperative_checkpoint.py

@@ -99,7 +99,7 @@ class MNIST(fluid.Layer):
 
 
 class TestDygraphCheckpoint(unittest.TestCase):
-    def save_load_persistables(self):
+    def test_save_load_persistables(self):
         seed = 90
         epoch_num = 1
 
@@ -135,23 +135,26 @@ class TestDygraphCheckpoint(unittest.TestCase):
 
                     avg_loss.backward()
                     sgd.minimize(avg_loss)
-                    fluid.dygraph.save_persistables(mnist, "save_dir")
+                    fluid.dygraph.save_persistables(mnist.state_dict(),
+                                                    "save_dir")
                     mnist.clear_gradients()
 
                     for param in mnist.parameters():
                         dy_param_init_value[param.name] = param.numpy()
 
                     mnist.load_dict(
-                        fluid.dygraph.load_persistables(mnist, "save_dir"))
+                        fluid.dygraph.load_persistables(mnist.state_dict(),
+                                                        "save_dir"))
 
                     restore = mnist.parameters()
 
                     self.assertEqual(len(dy_param_init_value), len(restore))
                     for value in restore:
                         self.assertTrue(
-                            np.allclose(value, dy_param_init_value[value.name]))
-                        self.assertTrue(np.isfinite(value.all()))
-                        self.assertFalse(np.isnan(value.any()))
+                            np.allclose(value.numpy(), dy_param_init_value[
+                                value.name]))
+                        self.assertTrue(np.isfinite(value.numpy().all()))
+                        self.assertFalse(np.isnan(value.numpy().any()))
 
                     step += 1