fix bug in boilerplate (#260)

* fix bug in boilerplate, self.cost and enhance FakeData

* fix len_range -> max_range

examples/boilerplate.py

@@ -13,19 +13,21 @@ All code is in this file is the most minimalistic way to solve a deep-learning p
 """
 
 BATCH_SIZE = 16
+SHAPE = 28
+CHANNELS = 3
 
 
 class Model(ModelDesc):
     def _get_inputs(self):
-        return [InputDesc(tf.float32, (None, 28, 28, 1), 'input'),
+        return [InputDesc(tf.float32, (None, SHAPE, SHAPE, CHANNELS), 'input'),
                 InputDesc(tf.int32, (None,), 'label')]
 
     def _build_graph(self, inputs):
         image, label = inputs
         image = image * 2 - 1
 
-        self.cost = tf.identity(0, name='total_costs')
-        summary.add_moving_summary(cost)
+        self.cost = tf.identity(0., name='total_costs')
+        summary.add_moving_summary(self.cost)
 
     def _get_optimizer(self):
         lr = symbolic_functions.get_scalar_var('learning_rate', 5e-3, summary=True)
@@ -33,9 +35,9 @@ class Model(ModelDesc):
 
 
 def get_data(subset):
-    ds = None  # something.get_data() that yields [[28, 28, 1], [1]]
-    # ...
-
+    # something that yields [[SHAPE, SHAPE, CHANNELS], [1]]
+    ds = FakeData([[SHAPE, SHAPE, CHANNELS], [1]], 1000, random=False,
+                  dtype=['float32', 'uint8'], domain=[(0, 255), (0, 10)])
     ds = PrefetchDataZMQ(ds, 2)
     ds = BatchData(ds, BATCH_SIZE)
     return ds

tensorpack/dataflow/raw.py

@@ -5,6 +5,7 @@
 
 import numpy as np
 import copy
+import six
 from six.moves import range
 from .base import DataFlow, RNGDataFlow
 
@@ -14,20 +15,22 @@ __all__ = ['FakeData', 'DataFromQueue', 'DataFromList']
 class FakeData(RNGDataFlow):
     """ Generate fake data of given shapes"""
 
-    def __init__(self, shapes, size=1000, random=True, dtype='float32'):
+    def __init__(self, shapes, size=1000, random=True, dtype='float32', domain=(0, 1)):
         """
         Args:
             shapes (list): a list of lists/tuples. Shapes of each component.
             size (int): size of this DataFlow.
             random (bool): whether to randomly generate data every iteration.
                 Note that merely generating the data could sometimes be time-consuming!
-            dtype (str): data type.
+            dtype (str): data type as string or a list of data types.
+            domain (str): domain of values as tuple/list.
         """
         super(FakeData, self).__init__()
         self.shapes = shapes
         self._size = int(size)
         self.random = random
-        self.dtype = dtype
+        self.dtype = [dtype] * len(shapes) if isinstance(dtype, six.string_types) else dtype
+        self.domain = [domain] * len(shapes) if isinstance(domain, tuple) else domain
 
     def size(self):
         return self._size
@@ -35,11 +38,18 @@ class FakeData(RNGDataFlow):
     def get_data(self):
         if self.random:
             for _ in range(self._size):
-                yield [self.rng.rand(*k).astype(self.dtype) for k in self.shapes]
+                val = []
+                for k in range(len(self.shapes)):
+                    v = self.rng.rand(*self.shapes[k]) * (self.domain[k][1] - self.domain[k][0]) + self.domain[k][0]
+                    val.append(v.astype(self.dtype[k]))
+                yield val
         else:
-            v = [self.rng.rand(*k).astype(self.dtype) for k in self.shapes]
+            val = []
+            for k in range(len(self.shapes)):
+                v = self.rng.rand(*self.shapes[k]) * (self.domain[k][1] - self.domain[k][0]) + self.domain[k][0]
+                val.append(v.astype(self.dtype[k]))
             for _ in range(self._size):
-                yield copy.deepcopy(v)
+                yield copy.deepcopy(val)
 
 
 class DataFromQueue(DataFlow):