[Fix] Rename init_weight to init_weights (#971)

* [Fix] Rename init_weight to init_weights

* warning msg

mmcv/runner/base_module.py

@@ -34,7 +34,7 @@ class BaseModule(nn.Module, metaclass=ABCMeta):
     def is_init(self):
         return self._is_init
 
-    def init_weight(self):
+    def init_weights(self):
         """Initialize the weights."""
         from ..cnn import initialize
 
@@ -42,11 +42,11 @@ class BaseModule(nn.Module, metaclass=ABCMeta):
             if self.init_cfg:
                 initialize(self, self.init_cfg)
             for m in self.children():
-                if hasattr(m, 'init_weight'):
-                    m.init_weight()
+                if hasattr(m, 'init_weights'):
+                    m.init_weights()
             self._is_init = True
         else:
-            warnings.warn(f'init_weight of {self.__class__.__name__} has '
+            warnings.warn(f'init_weights of {self.__class__.__name__} has '
                           f'been called more than once.')
 
     def __repr__(self):

tests/test_runner/test_basemodule.py

@@ -119,7 +119,7 @@ def test_model_weight_init():
             conv1d=dict(type='FooConv1d')))
 
     model = build_from_cfg(model_cfg, FOOMODELS)
-    model.init_weight()
+    model.init_weights()
 
     assert torch.equal(model.component1.conv1d.weight,
                        torch.full(model.component1.conv1d.weight.shape, 3.0))
@@ -199,7 +199,7 @@ def test_nest_components_weight_init():
             conv1d=dict(type='FooConv1d')))
 
     model = build_from_cfg(model_cfg, FOOMODELS)
-    model.init_weight()
+    model.init_weights()
 
     assert torch.equal(model.component1.conv1d.weight,
                        torch.full(model.component1.conv1d.weight.shape, 7.0))
@@ -243,7 +243,7 @@ def test_without_layer_weight_init():
         component2=dict(type='FooConv2d'),
         component3=dict(type='FooLinear'))
     model = build_from_cfg(model_cfg, FOOMODELS)
-    model.init_weight()
+    model.init_weights()
 
     assert torch.equal(model.component1.conv1d.weight,
                        torch.full(model.component1.conv1d.weight.shape, 3.0))
@@ -276,7 +276,7 @@ def test_override_weight_init():
         component1=dict(type='FooConv1d'),
         component3=dict(type='FooLinear'))
     model = build_from_cfg(model_cfg, FOOMODELS)
-    model.init_weight()
+    model.init_weights()
     assert torch.equal(model.reg.weight,
                        torch.full(model.reg.weight.shape, 10.0))
     assert torch.equal(model.reg.bias, torch.full(model.reg.bias.shape, 20.0))
@@ -308,7 +308,7 @@ def test_override_weight_init():
         component2=dict(type='FooConv2d'),
         component3=dict(type='FooLinear'))
     model = build_from_cfg(model_cfg, FOOMODELS)
-    model.init_weight()
+    model.init_weights()
 
     assert torch.equal(model.reg.weight,
                        torch.full(model.reg.weight.shape, 30.0))
@@ -326,7 +326,7 @@ def test_sequential_model_weight_init():
     ]
     layers = [build_from_cfg(cfg, COMPONENTS) for cfg in seq_model_cfg]
     seq_model = Sequential(*layers)
-    seq_model.init_weight()
+    seq_model.init_weights()
     assert torch.equal(seq_model[0].conv1d.weight,
                        torch.full(seq_model[0].conv1d.weight.shape, 0.))
     assert torch.equal(seq_model[0].conv1d.bias,
@@ -341,7 +341,7 @@ def test_sequential_model_weight_init():
         *layers,
         init_cfg=dict(
             type='Constant', layer=['Conv1d', 'Conv2d'], val=4., bias=5.))
-    seq_model.init_weight()
+    seq_model.init_weights()
     assert torch.equal(seq_model[0].conv1d.weight,
                        torch.full(seq_model[0].conv1d.weight.shape, 0.))
     assert torch.equal(seq_model[0].conv1d.bias,
@@ -363,7 +363,7 @@ def test_modulelist_weight_init():
     ]
     layers = [build_from_cfg(cfg, COMPONENTS) for cfg in models_cfg]
     modellist = ModuleList(layers)
-    modellist.init_weight()
+    modellist.init_weights()
     assert torch.equal(modellist[0].conv1d.weight,
                        torch.full(modellist[0].conv1d.weight.shape, 0.))
     assert torch.equal(modellist[0].conv1d.bias,
@@ -378,7 +378,7 @@ def test_modulelist_weight_init():
         layers,
         init_cfg=dict(
             type='Constant', layer=['Conv1d', 'Conv2d'], val=4., bias=5.))
-    modellist.init_weight()
+    modellist.init_weights()
     assert torch.equal(modellist[0].conv1d.weight,
                        torch.full(modellist[0].conv1d.weight.shape, 0.))
     assert torch.equal(modellist[0].conv1d.bias,