Python theano 模块，foldl() 实例源码

def foldl(fn, elems, initializer=None, name=None):
    '''Reduce elems using fn to combine them from left to right.

    # Arguments
        fn: Callable that will be called upon each element in elems and an
            accumulator, for instance lambda acc, x: acc + x
        elems: tensor
        initializer: The first value used (elems[0] in case of None)
        name: A string name for the foldl node in the graph

    # Returns
        Same type and shape as initializer
    '''
    if initializer is None:
        initializer = elems[0]
        elems = elems[1:]

    # We need to change the order of the arguments because theano accepts x as
    # first parameter and accumulator as second
    fn2 = lambda x, acc: fn(acc, x)

    return theano.foldl(fn2, elems, initializer, name=name)[0]

def foldl(fn, elems, initializer=None, name=None):
    """Reduce elems using fn to combine them from left to right.

    # Arguments
        fn: Callable that will be called upon each element in elems and an
            accumulator, for instance lambda acc, x: acc + x
        elems: tensor
        initializer: The first value used (elems[0] in case of None)
        name: A string name for the foldl node in the graph

    # Returns
        Same type and shape as initializer
    """
    if initializer is None:
        initializer = elems[0]
        elems = elems[1:]

    # We need to change the order of the arguments because theano accepts x as
    # first parameter and accumulator as second
    fn2 = lambda x, acc: fn(acc, x)

    return theano.foldl(fn2, elems, initializer, name=name)[0]

def foldl(fn, elems, initializer=None, name=None):
    '''Reduce elems using fn to combine them from left to right.

    # Arguments
        fn: Callable that will be called upon each element in elems and an
            accumulator, for instance lambda acc, x: acc + x
        elems: tensor
        initializer: The first value used (elems[0] in case of None)
        name: A string name for the foldl node in the graph

    # Returns
        Same type and shape as initializer
    '''
    if initializer is None:
        initializer = elems[0]
        elems = elems[1:]

    # We need to change the order of the arguments because theano accepts x as
    # first parameter and accumulator as second
    fn2 = lambda x, acc: fn(acc, x)

    return theano.foldl(fn2, elems, initializer, name=name)[0]

def foldl(fn, elems, initializer=None, name=None):
    """Reduce elems using fn to combine them from left to right.

    # Arguments
        fn: Callable that will be called upon each element in elems and an
            accumulator, for instance lambda acc, x: acc + x
        elems: tensor
        initializer: The first value used (elems[0] in case of None)
        name: A string name for the foldl node in the graph

    # Returns
        Same type and shape as initializer
    """
    if initializer is None:
        initializer = elems[0]
        elems = elems[1:]

    # We need to change the order of the arguments because theano accepts x as
    # first parameter and accumulator as second
    fn2 = lambda x, acc: fn(acc, x)

    return theano.foldl(fn2, elems, initializer, name=name)[0]

def foldl(fn, elems, initializer=None, name=None):
    """Reduce elems using fn to combine them from left to right.

    # Arguments
        fn: Callable that will be called upon each element in elems and an
            accumulator, for instance lambda acc, x: acc + x
        elems: tensor
        initializer: The first value used (elems[0] in case of None)
        name: A string name for the foldl node in the graph

    # Returns
        Same type and shape as initializer
    """
    if initializer is None:
        initializer = elems[0]
        elems = elems[1:]

    # We need to change the order of the arguments because theano accepts x as
    # first parameter and accumulator as second
    fn2 = lambda x, acc: fn(acc, x)

    return theano.foldl(fn2, elems, initializer, name=name)[0]

def accuracy_instance(predictions, targets, n=[1, 2, 3, 4, 5, 10], \
        nb_classes=5, nb_samples_per_class=10, batch_size=1):
    accuracy_0 = theano.shared(np.zeros((batch_size, nb_samples_per_class), \
        dtype=theano.config.floatX))
    indices_0 = theano.shared(np.zeros((batch_size, nb_classes), \
        dtype=np.int32))
    batch_range = T.arange(batch_size)
    def step_(p, t, acc, idx):
        acc = T.inc_subtensor(acc[batch_range, idx[batch_range, t]], T.eq(p, t))
        idx = T.inc_subtensor(idx[batch_range, t], 1)
        return (acc, idx)
    (raw_accuracy, _), _ = theano.foldl(step_, sequences=[predictions.dimshuffle(1, 0), \
        targets.dimshuffle(1, 0)], outputs_info=[accuracy_0, indices_0])
    accuracy = T.mean(raw_accuracy / nb_classes, axis=0)

    return accuracy

def test_foldl_memory_consumption(self):
        x = theano.shared(numpy.asarray(
            numpy.random.uniform(size=(10,)), dtype=theano.config.floatX))
        o, _ = theano.foldl(lambda v, acc: acc + v,
                            x,
                            theano.tensor.constant(
                                numpy.asarray(0.,
                                              dtype=theano.config.floatX)))

        mode = theano.compile.mode.FAST_RUN
        mode = mode.excluding('inplace')
        f0 = theano.function([], o, mode=mode)
        inputs, outputs = clone_optimized_graph(f0)

        scan_nodes = grab_scan_node(outputs[0])
        assert scan_nodes is not None
        scan_node = scan_nodes[0]
        f1 = theano.function(inputs, scan_node.inputs[2])

        # Originally, the shape would have been 1 due to the SaveMem
        # optimization reducing the size to the number of taps (in this case
        # 1) provided to the inner function. Now, because of the memory-reuse
        # feature in Scan it can be 2 because SaveMem needs to keep a
        # larger buffer to avoid aliasing between the inputs and the outputs.
        if theano.config.scan.allow_output_prealloc:
            assert f1().shape[0] == 2
        else:
            assert f1().shape[0] == 1

        gx = theano.tensor.grad(o, x)
        f2 = theano.function([], gx)
        utt.assert_allclose(f2(), numpy.ones((10,)))