Use int64 in reduction instead of casting to int

casting to int could lead to issues in the following case: - 64-bit Windows - python 2 - shape bigger than 2**31 In that case, int would be 32-bit, but shapes could be bigger and stored in long.

Use int64 in reduction instead of casting to int
4631b6ef · Pascal Lamblin · 83d3c3fb · 4631b6ef
--- a/theano/tensor/basic.py
+++ b/theano/tensor/basic.py
@@ -1233,9 +1233,9 @@ class MaxAndArgmax(Op):
        kept_shape = transposed_x.shape[:len(keep_axes)]
        reduced_shape = transposed_x.shape[len(keep_axes):]

-        # Numpy.prod returns 1.0 when arg is empty, so we cast it to int
-        # Otherwise reshape would complain citiing float arg
-        new_shape = kept_shape + (int(numpy.prod(reduced_shape)),)
+        # Numpy.prod returns 1.0 when arg is empty, so we cast it to int64
+        # Otherwise reshape would complain citing float arg
+        new_shape = kept_shape + (numpy.prod(reduced_shape, dtype='int64'),)
        reshaped_x = transposed_x.reshape(new_shape)

        max_idx[0] = theano._asarray(numpy.argmax(reshaped_x, axis=-1),