本文整理汇总了Python中blocks.bricks.NDimensionalSoftmax.name方法的典型用法代码示例。如果您正苦于以下问题：Python NDimensionalSoftmax.name方法的具体用法？Python NDimensionalSoftmax.name怎么用？Python NDimensionalSoftmax.name使用的例子？那么, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类blocks.bricks.NDimensionalSoftmax的用法示例。
在下文中一共展示了NDimensionalSoftmax.name方法的1个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: SimpleSpeechRecognizer
# 需要导入模块: from blocks.bricks import NDimensionalSoftmax [as 别名]
# 或者: from blocks.bricks.NDimensionalSoftmax import name [as 别名]
# ******************* Model *******************
recognizer = SimpleSpeechRecognizer(transition=transition,
dims_transition=conf.dims_transition,
num_features=num_features, num_classes=num_classes)
#recognizer = SpeechRecognizer(
# num_features=num_features, dims_bottom=[],
# dims_bidir=conf.dims_transition, dims_top=[num_classes],
# bidir_trans=GatedRecurrent, bottom_activation=None)
# ******************* output *******************
y_hat = recognizer.apply(x,x_m)
y_hat.name = 'outputs'
y_hat_softmax = NDimensionalSoftmax().apply(y_hat, extra_ndim = y_hat.ndim - 2)
y_hat_softmax.name = 'outputs_softmax'
# there is a cost function for monitoring and for training, because one is more stable to compute
# gradients and seems also to be more memory efficient, but does not compute the true cost.
if conf.task=='CTC':
cost_train = ctc.pseudo_cost(y, y_hat, y_m, x_m).mean()
cost_train.name = "cost_train"
cost_monitor = ctc.cost(y, y_hat_softmax, y_m, x_m).mean()
cost_monitor.name = "cost_monitor"
elif conf.task=='framewise':
cost_train = categorical_crossentropy_batch().apply(y_hat_softmax, y, x_m)
cost_train.name='cost'
cost_monitor = cost_train
else:
raise ValueError, conf.task