Fix use of InputDesc across graphs (fix #398)

docs/tutorial/callback.md

@@ -15,10 +15,12 @@ There are several places where you might want to do something else:
 * After the training (e.g. send the model somewhere, send a message to your phone)
 
 We found people traditionally tend to write the training loop together with these extra features.
-This makes the loop lengthy, and the code for the same feature probably get separated.
+This makes the loop lengthy, and the code for the same feature probably get separated (imagine a
+feature which needs initialization in the beginning and then some actual work between iterations).
+
 By writing callbacks to implement what to do at each place, tensorpack trainers
 will call the callbacks at the proper time.
-Therefore the code can be reused with one single line, as long as you are using tensorpack trainers.
+Therefore these features can be reused with one single line, as long as you are using tensorpack trainers.
 
 For example, these are the callbacks I used when training a ResNet:
 
@@ -30,7 +32,7 @@ TrainConfig(
     ModelSaver(),
     # backup the model with best validation error
     MinSaver('val-error-top1'),
-    # run inference on another Dataflow every epoch, compute top1/top5 classification error and save them in log
+    # run inference on another Dataflow every epoch, compute classification error and log to monitors
     InferenceRunner(dataset_val, [
         ClassificationError('wrong-top1', 'val-error-top1'),
         ClassificationError('wrong-top5', 'val-error-top5')]),
@@ -50,11 +52,11 @@ TrainConfig(
     InjectShell(shell='ipython')
   ],
   extra_callbacks=[    # these callbacks are enabled by default already
-    # maintain and summarize moving average of some tensors defined in the model (e.g. training loss, training error)
+    # maintain those moving average summaries already defined in the model (e.g. training loss, training error)
     MovingAverageSummary(),
     # draw a nice progress bar
     ProgressBar(),
-    # run `tf.summary.merge_all` every epoch and send results to monitors
+    # run `tf.summary.merge_all` every epoch and log to monitors
     MergeAllSummaries(),
     # run ops in GraphKeys.UPDATE_OPS collection along with training, if any
     RunUpdateOps(),

tensorpack/graph_builder/model_desc.py

@@ -25,8 +25,6 @@ class InputDesc(
     input source.
     """
 
-    _cached_placeholder = None
-
     def __new__(cls, type, shape, name):
         """
         Args:
@@ -36,6 +34,7 @@ class InputDesc(
         """
         shape = tuple(shape)    # has to be tuple for self to be hashable
         self = super(InputDesc, cls).__new__(cls, type, shape, name)
+        self._cached_placeholder = None
         return self
 
     # TODO in serialization, skip _cached_placeholder
@@ -72,10 +71,10 @@ class InputDesc(
                 self.type, shape=self.shape,
                 name=prefix + self.name)
         if prefix == '' and self._cached_placeholder is None:
-            self._cached_placeholder = ret
+            self._cached_placeholder = ret  # cached_placeholder only caches the prefix='' case
         return ret
 
-    @memoized
+    # cannot memoize here, because InputDesc is hashed by its fields.
     def build_placeholder_reuse(self):
         """
         Build a tf.placeholder from the metadata, or return an old one.