support numpy dot semantic
>>> w = T.matrix()
>>> m = T.tensor3()
>>> d = T.dot(m, w)
Traceback (most recent call last):
File '', line 1, in
File '/Volumes/Data/src/varia/theano/theano/gof/op.py', line 323, in +call+
node = self.make_node(* inputs, **kwargs)
File '/Volumes/Data/src/varia/theano/theano/tensor/basic.py', line
4101, in make_node
' enabling numpy dot semantics if you want them'), x)
TypeError: ('dot supports matrix and vector args: email theano-dev
about enabling numpy dot semantics if you want
them',
)
>>> m = T.tensor3()
>>> d = T.dot(m, w)
Traceback (most recent call last):
File '', line 1, in
File '/Volumes/Data/src/varia/theano/theano/gof/op.py', line 323, in +call+
node = self.make_node(* inputs, **kwargs)
File '/Volumes/Data/src/varia/theano/theano/tensor/basic.py', line
4101, in make_node
' enabling numpy dot semantics if you want them'), x)
TypeError: ('dot supports matrix and vector args: email theano-dev
about enabling numpy dot semantics if you want
them',
)
Leave a comment
There is a keeping "numpy_semantics" attribute or param that shoud enable this. But this is not tested.
Also, we don't konw the full implication w/ respect to efficiency. What is sure is that it will use NumPy binding to blas, not ours. I don't know how well they use BLAS in that case.
Also, we don't konw the full implication w/ respect to efficiency. What is sure is that it will use NumPy binding to blas, not ours. I don't know how well they use BLAS in that case.