[Fix] Delete frozen parameters when using paramwise_cfg (open-mmlab…

…#1441)
fanqiNO1 · Apr 22, 2024 · acbc5e4 · acbc5e4
1 parent 9ecced8
commit acbc5e4
Show file tree

Hide file tree

Showing 2 changed files with 15 additions and 20 deletions.
diff --git a/mmengine/optim/optimizer/default_constructor.py b/mmengine/optim/optimizer/default_constructor.py
@@ -213,7 +213,10 @@ def add_params(self,
                     level=logging.WARNING)
                 continue
             if not param.requires_grad:
-                params.append(param_group)
+                print_log((f'{prefix}.{name} is skipped since its '
+                           f'requires_grad={param.requires_grad}'),
+                          logger='current',
+                          level=logging.WARNING)
                 continue
 
             # if the parameter match one of the custom keys, ignore other rules

diff --git a/tests/test_optim/test_optimizer/test_optimizer.py b/tests/test_optim/test_optimizer/test_optimizer.py
@@ -549,7 +549,8 @@ def test_default_optimizer_constructor_with_empty_paramwise_cfg(self):
                 weight_decay=self.base_wd,
                 momentum=self.momentum))
         paramwise_cfg = dict()
-        optim_constructor = DefaultOptimWrapperConstructor(optim_wrapper_cfg)
+        optim_constructor = DefaultOptimWrapperConstructor(
+            optim_wrapper_cfg, paramwise_cfg)
         optim_wrapper = optim_constructor(model)
         self._check_default_optimizer(optim_wrapper.optimizer, model)
 
@@ -591,23 +592,16 @@ def test_default_optimizer_constructor_no_grad(self):
             dwconv_decay_mult=0.1,
             dcn_offset_lr_mult=0.1)
 
-        for param in self.model.parameters():
-            param.requires_grad = False
+        self.model.conv1.requires_grad_(False)
         optim_constructor = DefaultOptimWrapperConstructor(
             optim_wrapper_cfg, paramwise_cfg)
         optim_wrapper = optim_constructor(self.model)
-        optimizer = optim_wrapper.optimizer
-        param_groups = optimizer.param_groups
-        assert isinstance(optim_wrapper.optimizer, torch.optim.SGD)
-        assert optimizer.defaults['lr'] == self.base_lr
-        assert optimizer.defaults['momentum'] == self.momentum
-        assert optimizer.defaults['weight_decay'] == self.base_wd
-        for i, (name, param) in enumerate(self.model.named_parameters()):
-            param_group = param_groups[i]
-            assert torch.equal(param_group['params'][0], param)
-            assert param_group['momentum'] == self.momentum
-            assert param_group['lr'] == self.base_lr
-            assert param_group['weight_decay'] == self.base_wd
+
+        all_params = []
+        for pg in optim_wrapper.param_groups:
+            all_params.extend(map(id, pg['params']))
+        self.assertNotIn(id(self.model.conv1.weight), all_params)
+        self.assertIn(id(self.model.conv2.weight), all_params)
 
     def test_default_optimizer_constructor_bypass_duplicate(self):
         # paramwise_cfg with bypass_duplicate option
@@ -663,10 +657,8 @@ def test_default_optimizer_constructor_bypass_duplicate(self):
         optim_wrapper = optim_constructor(model)
         model_parameters = list(model.parameters())
         num_params = 14 if MMCV_FULL_AVAILABLE else 11
-        assert len(optim_wrapper.optimizer.param_groups) == len(
-            model_parameters) == num_params
-        self._check_sgd_optimizer(optim_wrapper.optimizer, model,
-                                  **paramwise_cfg)
+        assert len(optim_wrapper.optimizer.param_groups
+                   ) == len(model_parameters) - 1 == num_params - 1
 
     def test_default_optimizer_constructor_custom_key(self):
         # test DefaultOptimWrapperConstructor with custom_keys and