當前位置: 首頁>>代碼示例>>Python>>正文

Python recurrent.SimpleRecurrent方法代碼示例

本文整理匯總了Python中blocks.bricks.recurrent.SimpleRecurrent方法的典型用法代碼示例。如果您正苦於以下問題：Python recurrent.SimpleRecurrent方法的具體用法？Python recurrent.SimpleRecurrent怎麽用？Python recurrent.SimpleRecurrent使用的例子？那麽, 這裏精選的方法代碼示例或許可以為您提供幫助。您也可以進一步了解該方法所在類blocks.bricks.recurrent的用法示例。


在下文中一共展示了recurrent.SimpleRecurrent方法的3個代碼示例，這些例子默認根據受歡迎程度排序。您可以為喜歡或者感覺有用的代碼點讚，您的評價將有助於係統推薦出更棒的Python代碼示例。

示例1: __init__

# 需要導入模塊: from blocks.bricks import recurrent [as 別名]
# 或者: from blocks.bricks.recurrent import SimpleRecurrent [as 別名]
def __init__(self, dimension, alphabet_size, **kwargs):
        super(SimpleGenerator, self).__init__(**kwargs)
        lookup = LookupTable(alphabet_size, dimension)
        transition = SimpleRecurrent(
            activation=Tanh(),
            dim=dimension, name="transition")
        attention = SequenceContentAttention(
            state_names=transition.apply.states,
            attended_dim=dimension, match_dim=dimension, name="attention")
        readout = Readout(
            readout_dim=alphabet_size,
            source_names=[transition.apply.states[0],
                          attention.take_glimpses.outputs[0]],
            emitter=SoftmaxEmitter(name="emitter"),
            feedback_brick=LookupFeedback(alphabet_size, dimension),
            name="readout")
        generator = SequenceGenerator(
            readout=readout, transition=transition, attention=attention,
            name="generator")

        self.lookup = lookup
        self.generator = generator
        self.children = [lookup, generator]
開發者ID:rizar，項目名稱:attention-lvcsr，代碼行數:25，代碼來源:test_search.py

示例2: rnn_layer

# 需要導入模塊: from blocks.bricks import recurrent [as 別名]
# 或者: from blocks.bricks.recurrent import SimpleRecurrent [as 別名]
def rnn_layer(dim, h, n):
    linear = Linear(input_dim=dim, output_dim=dim, name='linear' + str(n))
    rnn = SimpleRecurrent(dim=dim, activation=Tanh(), name='rnn' + str(n))
    initialize([linear, rnn])
    return rnn.apply(linear.apply(h))
開發者ID:johnarevalo，項目名稱:blocks-char-rnn，代碼行數:7，代碼來源:model.py

示例3: __init__

# 需要導入模塊: from blocks.bricks import recurrent [as 別名]
# 或者: from blocks.bricks.recurrent import SimpleRecurrent [as 別名]
def __init__(self, dimension, alphabet_size, **kwargs):
        super(WordReverser, self).__init__(**kwargs)
        encoder = Bidirectional(
            SimpleRecurrent(dim=dimension, activation=Tanh()))
        fork = Fork([name for name in encoder.prototype.apply.sequences
                    if name != 'mask'])
        fork.input_dim = dimension
        fork.output_dims = [encoder.prototype.get_dim(name) for name in fork.input_names]
        lookup = LookupTable(alphabet_size, dimension)
        transition = SimpleRecurrent(
            activation=Tanh(),
            dim=dimension, name="transition")
        attention = SequenceContentAttention(
            state_names=transition.apply.states,
            attended_dim=2 * dimension, match_dim=dimension, name="attention")
        readout = Readout(
            readout_dim=alphabet_size,
            source_names=[transition.apply.states[0],
                          attention.take_glimpses.outputs[0]],
            emitter=SoftmaxEmitter(name="emitter"),
            feedback_brick=LookupFeedback(alphabet_size, dimension),
            name="readout")
        generator = SequenceGenerator(
            readout=readout, transition=transition, attention=attention,
            name="generator")

        self.lookup = lookup
        self.fork = fork
        self.encoder = encoder
        self.generator = generator
        self.children = [lookup, fork, encoder, generator]
開發者ID:mila-iqia，項目名稱:blocks-examples，代碼行數:33，代碼來源:__init__.py


注：本文中的blocks.bricks.recurrent.SimpleRecurrent方法示例由純淨天空整理自Github/MSDocs等開源代碼及文檔管理平台，相關代碼片段篩選自各路編程大神貢獻的開源項目，源碼版權歸原作者所有，傳播和使用請參考對應項目的License；未經允許，請勿轉載。