docs update

docs/conf.py

@@ -363,6 +363,7 @@ def autodoc_skip_member(app, what, name, obj, skip, options):
         'predictor_factory',
         'get_predictors',
         'vs_name_for_predictor',
+        'RandomCropAroundBox',
         'dump_chkpt_vars',
         'VisualQA',
         'ParamRestore']:

docs/tutorial/graph.md

@@ -32,19 +32,20 @@ the argument `inputs` is the list of input tensors matching `_get_inputs`.
 You can use any symbolic functions in `_build_graph`, including TensorFlow core library
 functions and other symbolic libraries.
 
-### How is it Used:
+### How it is Used:
 
 Most tensorpack trainers expect a `ModelDesc`, and use it as a __description
-of the TF graph__ (except for the input pipeline).
+of the TF graph to be built__.
 These trainers will use `_get_inputs` to connect the given `InputSource` to the graph.
 They'll then use `_build_graph` to create the backbone model, and then `_get_optimizer` to create the minimization op, and run it.
 
 Note that data-parallel multi-GPU trainers will call `_build_graph` __multiple times__ on each GPU.
 A trainer may also make __extra calls__ to `_build_graph` for inference, if used by some callbacks.
-`_build_graph` will always be called under some `TowerContext` which contains these information
+`_build_graph` will always be called under some `TowerContext` which contains these context information
 (e.g. training or inference, reuse or not, scope name) for your access.
 
-Also, to respect variable reuse among multiple calls, use `tf.get_variable()` instead of `tf.Variable`.
+Also, to respect variable reuse among multiple calls, use `tf.get_variable()` instead of `tf.Variable` in `_build_graph`,
+if you need to create and variables.
 
 ### Build It Manually
 

tensorpack/dataflow/imgaug/crop.py

@@ -9,8 +9,7 @@ from ...utils.argtools import shape2d
 from six.moves import range
 import numpy as np
 
-__all__ = ['RandomCrop', 'CenterCrop',
-           'perturb_BB', 'RandomCropAroundBox', 'RandomCropRandomShape']
+__all__ = ['RandomCrop', 'CenterCrop', 'RandomCropAroundBox', 'RandomCropRandomShape']
 
 
 class RandomCrop(ImageAugmentor):
@@ -111,9 +110,10 @@ def perturb_BB(image_shape, bb, max_perturb_pixel,
     return bb
 
 
+# TODO shouldn't include strange augmentors like this.
 class RandomCropAroundBox(ImageAugmentor):
     """
-    Crop a box around a bounding box by some random perturbation
+    Crop a box around a bounding box by some random perturbation.
     """
 
     def __init__(self, perturb_ratio, max_aspect_ratio_diff=0.3):

tensorpack/dataflow/prefetch.py

@@ -150,7 +150,7 @@ class PrefetchDataZMQ(ProxyDataFlow):
             nr_proc (int): number of processes to use.
             hwm (int): the zmq "high-water mark" for both sender and receiver.
         """
-        assert os.name != 'nt', "PrefetchDataZMQ doesn't support windows! Consider PrefetchData instead."
+        assert os.name != 'nt', "PrefetchDataZMQ doesn't support windows!  PrefetchData might work sometimes."
         super(PrefetchDataZMQ, self).__init__(ds)
         try:
             self._size = ds.size()