In ShapeFeature, use a different variable for each env...

theano/tensor/opt.py

@@ -491,8 +491,6 @@ class Shape_i(T.Op):
     def grad(self, inp, grads):
         return [None]
 
-lscalar_one = T.constant(1, dtype='int64')
-assert lscalar_one.type == T.lscalar
 class ShapeFeature(object):
     """Graph optimizer for removing all calls to shape()
 
@@ -568,12 +566,12 @@ class ShapeFeature(object):
     sometimes Theano constants?? That would be confusing.
 
     """
-    @staticmethod
-    def shape_ir(i, r):
+
+    def shape_ir(self, i, r):
         #TODO: Write a doc string for this method
 
         if hasattr(r.type,"broadcastable") and r.type.broadcastable[i]:
-            return lscalar_one
+            return self.lscalar_one
         else:
             return Shape_i(i).make_node(r).outputs[0]
 
@@ -595,7 +593,7 @@ class ShapeFeature(object):
         assert s_i is not None
         if s_i == 1:
             # don't make the optimizer merge a zillion ones together
-            return lscalar_one
+            return self.lscalar_one
         if type(s_i) is int or isinstance(s_i, numpy.integer):
             # this shape is a constant
             assert s_i >= 0
@@ -693,6 +691,11 @@ class ShapeFeature(object):
     def on_attach(self, env):
         assert not hasattr(env, 'shape_feature')
         env.shape_feature = self
+        # Must be local to the object as otherwise we reuse the same
+        # variable for multiple env!
+        self.lscalar_one = T.constant(1, dtype='int64')
+        assert self.lscalar_one.type == T.lscalar
+
         self.shape_of = {} # Variable -> tuple(scalars) or None  (All tensor vars map to tuple)
         self.scheduled = {} # Variable ->
         for node in env.toposort():