make optimizer an optional argument for Runner

mmcv/torchpack/runner/runner.py

@@ -20,13 +20,16 @@ class Runner(object):
 
     def __init__(self,
                  model,
-                 optimizer,
                  batch_processor,
+                 optimizer=None,
                  work_dir=None,
                  log_level=logging.INFO):
         assert callable(batch_processor)
         self.model = model
-        self.optimizer = self.init_optimizer(optimizer)
+        if optimizer is not None:
+            self.optimizer = self.init_optimizer(optimizer)
+        else:
+            self.optimizer = None
         self.batch_processor = batch_processor
 
         # create work_dir
@@ -152,6 +155,9 @@ class Runner(object):
         Returns:
             list: Current learning rate of all param groups.
         """
+        if self.optimizer is None:
+            raise RuntimeError(
+                'lr is not applicable because optimizer does not exist.')
         return [group['lr'] for group in self.optimizer.param_groups]
 
     def register_hook(self, hook, priority=50):
@@ -234,8 +240,9 @@ class Runner(object):
         for i, data_batch in enumerate(data_loader):
             self._inner_iter = i
             self.call_hook('before_val_iter')
-            outputs = self.batch_processor(
-                self.model, data_batch, train_mode=False, **kwargs)
+            with torch.no_grad():
+                outputs = self.batch_processor(
+                    self.model, data_batch, train_mode=False, **kwargs)
             if not isinstance(outputs, dict):
                 raise TypeError('batch_processor() must return a dict')
             if 'log_vars' in outputs:
@@ -321,12 +328,12 @@ class Runner(object):
                 info, hooks, default_args=dict(interval=log_interval))
             self.register_hook(logger_hook, priority=60)
 
-    def register_default_hooks(self,
-                               lr_config,
-                               grad_clip_config=None,
-                               checkpoint_config=None,
-                               log_config=None):
-        """Register several default hooks.
+    def register_training_hooks(self,
+                                lr_config,
+                                grad_clip_config=None,
+                                checkpoint_config=None,
+                                log_config=None):
+        """Register default hooks for training.
 
         Default hooks include:
         - LrUpdaterHook