some docs improvement about multi-gpu (fix #1084)

docs/tutorial/trainer.md

@@ -83,7 +83,9 @@ Note some __common problems__ when using these trainers:
 
 1. In each iteration, instead of taking one input tensor for all GPUs and split,
     all GPUs take tensors from the `InputSource`.
-	So the total batch size across all GPUs would become ``(batch size of InputSource) * #GPU``.
+	So the total batch size across all GPUs is ``(batch size of InputSource) * #GPU``.
+    You may want to change `steps_per_epoch` or learing rate appropriately according
+    to the total batch size.
 
     ```eval_rst
     .. note:: 
@@ -96,7 +98,7 @@ Note some __common problems__ when using these trainers:
     ```
 
 2. The tower function (your model code) will get called once on each GPU.
-   You must follow the abovementieond rules of tower function.
+   You must follow the abovementioned rules of tower function.
 
 ### Distributed Trainers
 

tensorpack/train/config.py

@@ -93,7 +93,7 @@ class TrainConfig(object):
 
             starting_epoch (int): The index of the first epoch.
             steps_per_epoch (int): the number of steps (defined by :meth:`Trainer.run_step`) to run in each epoch.
-                Defaults to the input data size.
+                Defaults to the input data size. You may want to divide it by the #GPUs in multi-GPU training.
             max_epoch (int): maximum number of epoch to run training.
         """
 
@@ -156,14 +156,6 @@ class TrainConfig(object):
         self.starting_epoch = int(starting_epoch)
         self.max_epoch = int(max_epoch)
 
-        if 'nr_tower' in kwargs:
-            self.nr_tower = kwargs.pop('nr_tower')
-        if 'tower' in kwargs:
-            self.tower = kwargs.pop('tower')
-        else:
-            self.tower = [0]
-        assert len(kwargs) == 0, "Unknown arguments: {}".format(kwargs.keys())
-
 
 class AutoResumeTrainConfig(TrainConfig):
     """