in {zeros,ones}_like, don't insert fill for scalar.

This make the graph before compilation and when compiled with FAST_COMPILE more readable. This was optimized away to local_fill_to_alloc, but if we enable it in fast_compile, it also do more optimisation that make the graph not readable without enabling even more optimization in fast_compile.s

in {zeros,ones}_like, don't insert fill for scalar.
187603d6 · Frederic · 5cd26de6 · 187603d6
--- a/theano/tensor/basic.py
+++ b/theano/tensor/basic.py
@@ -2650,6 +2650,8 @@ def ones_like(model, dtype=None):
    """equivalent of numpy.ones_like"""
    if dtype is None:
        dtype = model.type.dtype
+    if model.ndim == 0 and isinstance(model.type, TensorType):
+        return constant(1.0, dtype=dtype)
    ret = fill(model, constant(1.0, dtype=dtype))
    return ret

@@ -2659,6 +2661,8 @@ def zeros_like(model, dtype=None):
    """equivalent of numpy.zeros_like"""
    if dtype is None:
        dtype = model.type.dtype
+    if model.ndim == 0 and isinstance(model.type, TensorType):
+        return constant(0.0, dtype=dtype)
    return fill(model, constant(0.0, dtype=dtype))