align minibatches (#2760)

* align minibatches

* add simple test

* align specific minibatches

Author: ferrine

pymc3/data.py

@@ -2,7 +2,7 @@
 import io
 import os
 import pkgutil
-
+import collections
 import numpy as np
 import pymc3 as pm
 import theano.tensor as tt
@@ -11,7 +11,8 @@
 __all__ = [
     'get_data',
     'GeneratorAdapter',
-    'Minibatch'
+    'Minibatch',
+    'align_minibatches'
 ]
 
 
@@ -221,6 +222,9 @@ class Minibatch(tt.TensorVariable):
     >>> x = Minibatch(moredata, [2, None, 20, Ellipsis, 10])
     >>> assert x.eval().shape == (2, 20, 20, 40, 10)
     """
+
+    RNG = collections.defaultdict(list)
+
     @theano.configparser.change_flags(compute_test_value='raise')
     def __init__(self, data, batch_size=128, dtype=None, broadcastable=None, name='Minibatch',
                  random_seed=42, update_shared_f=None, in_memory_size=None):
@@ -244,17 +248,21 @@ def __init__(self, data, batch_size=128, dtype=None, broadcastable=None, name='M
             inputs=[self.minibatch], outputs=[self])
         self.tag.test_value = copy(self.minibatch.tag.test_value)
 
-    @staticmethod
-    def rslice(total, size, seed):
+    def rslice(self, total, size, seed):
         if size is None:
             return slice(None)
         elif isinstance(size, int):
-            return (pm.tt_rng(seed)
+            rng = pm.tt_rng(seed)
+            Minibatch.RNG[id(self)].append(rng)
+            return (rng
                     .uniform(size=(size, ), low=0.0, high=pm.floatX(total) - 1e-16)
                     .astype('int64'))
         else:
             raise TypeError('Unrecognized size type, %r' % size)
 
+    def __del__(self):
+        del Minibatch.RNG[id(self)]
+
     @staticmethod
     def make_static_slices(user_size):
         if user_size is None:
@@ -278,12 +286,11 @@ def make_static_slices(user_size):
         else:
             raise TypeError('Unrecognized size type, %r' % user_size)
 
-    @classmethod
-    def make_random_slices(cls, in_memory_shape, batch_size, default_random_seed):
+    def make_random_slices(self, in_memory_shape, batch_size, default_random_seed):
         if batch_size is None:
             return [Ellipsis]
         elif isinstance(batch_size, int):
-            slc = [cls.rslice(in_memory_shape[0], batch_size, default_random_seed)]
+            slc = [self.rslice(in_memory_shape[0], batch_size, default_random_seed)]
         elif isinstance(batch_size, (list, tuple)):
             def check(t):
                 if t is Ellipsis or t is None:
@@ -334,10 +341,10 @@ def check(t):
             else:
                 shp_end = np.asarray([])
             shp_begin = shape[:len(begin)]
-            slc_begin = [cls.rslice(shp_begin[i], t[0], t[1])
+            slc_begin = [self.rslice(shp_begin[i], t[0], t[1])
                          if t is not None else tt.arange(shp_begin[i])
                          for i, t in enumerate(begin)]
-            slc_end = [cls.rslice(shp_end[i], t[0], t[1])
+            slc_end = [self.rslice(shp_end[i], t[0], t[1])
                        if t is not None else tt.arange(shp_end[i])
                        for i, t in enumerate(end)]
             slc = slc_begin + mid + slc_end
@@ -359,3 +366,16 @@ def clone(self):
         ret.name = self.name
         ret.tag = copy(self.tag)
         return ret
+
+
+def align_minibatches(batches=None):
+    if batches is None:
+        for rngs in Minibatch.RNG.values():
+            for rng in rngs:
+                rng.seed()
+    else:
+        for b in batches:
+            if not isinstance(b, Minibatch):
+                raise TypeError('{b} is not a Minibatch')
+            for rng in Minibatch.RNG[id(b)]:
+                rng.seed()

pymc3/tests/test_minibatches.py

@@ -313,3 +313,21 @@ def test_cloning_available(self):
         res1 = theano.clone(res, {gop: shared})
         f = theano.function([], res1)
         assert f() == np.array([100])
+
+    def test_align(self):
+        m = pm.Minibatch(np.arange(1000), 1, random_seed=1)
+        n = pm.Minibatch(np.arange(1000), 1, random_seed=1)
+        f = theano.function([], [m, n])
+        n.eval()  # not aligned
+        a, b = zip(*(f() for _ in range(1000)))
+        assert a != b
+        pm.align_minibatches()
+        a, b = zip(*(f() for _ in range(1000)))
+        assert a == b
+        n.eval()  # not aligned
+        pm.align_minibatches([m])
+        a, b = zip(*(f() for _ in range(1000)))
+        assert a != b
+        pm.align_minibatches([m, n])
+        a, b = zip(*(f() for _ in range(1000)))
+        assert a == b