当前位置: 首页>>代码示例>>Python>>正文


Python ZoomableAttentionWindow.read方法代码示例

本文整理汇总了Python中attention.ZoomableAttentionWindow.read方法的典型用法代码示例。如果您正苦于以下问题:Python ZoomableAttentionWindow.read方法的具体用法?Python ZoomableAttentionWindow.read怎么用?Python ZoomableAttentionWindow.read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attention.ZoomableAttentionWindow的用法示例。


在下文中一共展示了ZoomableAttentionWindow.read方法的4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: AttentionReader

# 需要导入模块: from attention import ZoomableAttentionWindow [as 别名]
# 或者: from attention.ZoomableAttentionWindow import read [as 别名]
class AttentionReader(Initializable):
    def __init__(self, x_dim, dec_dim, channels, height, width, N, **kwargs):
        super(AttentionReader, self).__init__(name="reader", **kwargs)

        self.img_height = height
        self.img_width = width
        self.N = N
        self.x_dim = x_dim
        self.dec_dim = dec_dim
        self.output_dim = 2*channels*N*N

        self.zoomer = ZoomableAttentionWindow(channels, height, width, N)
        self.readout = MLP(activations=[Identity()], dims=[dec_dim, 5], **kwargs)

        self.children = [self.readout]

    def get_dim(self, name):
        if name == 'input':
            return self.dec_dim
        elif name == 'x_dim':
            return self.x_dim
        elif name == 'output':
            return self.output_dim
        else:
            raise ValueError
            
    @application(inputs=['x', 'x_hat', 'h_dec'], outputs=['r'])
    def apply(self, x, x_hat, h_dec):
        l = self.readout.apply(h_dec)

        center_y, center_x, delta, sigma, gamma = self.zoomer.nn2att(l)

        w     = gamma * self.zoomer.read(x    , center_y, center_x, delta, sigma)
        w_hat = gamma * self.zoomer.read(x_hat, center_y, center_x, delta, sigma)
        
        return T.concatenate([w, w_hat], axis=1)

    @application(inputs=['x', 'x_hat', 'h_dec'], outputs=['r','center_y', 'center_x', 'delta'])
    def apply_detailed(self, x, x_hat, h_dec):
        l = self.readout.apply(h_dec)

        center_y, center_x, delta, sigma, gamma = self.zoomer.nn2att(l)

        w     = gamma * self.zoomer.read(x    , center_y, center_x, delta, sigma)
        w_hat = gamma * self.zoomer.read(x_hat, center_y, center_x, delta, sigma)
        
        r = T.concatenate([w, w_hat], axis=1)
        return r, center_y, center_x, delta

    @application(inputs=['x', 'h_dec'], outputs=['r','center_y', 'center_x', 'delta'])
    def apply_simple(self, x, h_dec):
        l = self.readout.apply(h_dec)

        center_y, center_x, delta, sigma, gamma = self.zoomer.nn2att(l)

        r     = gamma * self.zoomer.read(x    , center_y, center_x, delta, sigma)

        return r, center_y, center_x, delta
开发者ID:drewlinsley,项目名称:draw_classify,代码行数:60,代码来源:draw_CL_WORKING.py

示例2: AttentionReader

# 需要导入模块: from attention import ZoomableAttentionWindow [as 别名]
# 或者: from attention.ZoomableAttentionWindow import read [as 别名]
class AttentionReader(Initializable):
    def __init__(self, x_dim, dec_dim, height, width, N, **kwargs):
        super(AttentionReader, self).__init__(name="reader", **kwargs)

        self.img_height = height
        self.img_width = width
        self.N = N
        self.x_dim = x_dim
        self.dec_dim = dec_dim
        self.output_dim = 2 * N * N

        self.zoomer = ZoomableAttentionWindow(height, width, N)
        self.readout = MLP(activations=[Identity()], dims=[dec_dim, 5], **kwargs)

        self.children = [self.readout]

    def get_dim(self, name):
        if name == "input":
            return self.dec_dim
        elif name == "x_dim":
            return self.x_dim
        elif name == "output":
            return self.output_dim
        else:
            raise ValueError

    @application(inputs=["x", "x_hat", "h_dec"], outputs=["r"])
    def apply(self, x, x_hat, h_dec):
        l = self.readout.apply(h_dec)

        center_y, center_x, delta, sigma, gamma = self.zoomer.nn2att(l)

        w = gamma * self.zoomer.read(x, center_y, center_x, delta, sigma)
        w_hat = gamma * self.zoomer.read(x_hat, center_y, center_x, delta, sigma)

        return T.concatenate([w, w_hat], axis=1)
开发者ID:zan2434,项目名称:draw,代码行数:38,代码来源:draw.py

示例3: LocatorReader

# 需要导入模块: from attention import ZoomableAttentionWindow [as 别名]
# 或者: from attention.ZoomableAttentionWindow import read [as 别名]
class LocatorReader(Initializable):
    def __init__(self, x_dim, dec_dim, channels, height, width, N, **kwargs):
        super(LocatorReader, self).__init__(name="reader", **kwargs)

        self.img_height = height
        self.img_width = width
        self.N = N
        self.x_dim = x_dim
        self.dec_dim = dec_dim
        self.output_dim = channels * N * N

        self.zoomer = ZoomableAttentionWindow(channels, height, width, N)
        self.readout = MLP(activations=[Identity()], dims=[dec_dim, 7], **kwargs)

        self.children = [self.readout]

    def get_dim(self, name):
        if name == 'input':
            return self.dec_dim
        elif name == 'x_dim':
            return self.x_dim
        elif name == 'output':
            return self.output_dim
        else:
            raise ValueError

    @application(inputs=['x', 'h_dec'], outputs=['r', 'l'])
    def apply(self, x, h_dec):
        l = self.readout.apply(h_dec)

        center_y, center_x, deltaY, deltaX, sigmaY, sigmaX, gamma = self.zoomer.nn2att(l)

        w = gamma * self.zoomer.read(x, center_y, center_x, deltaY, deltaX, sigmaY, sigmaX)

        return w, l

    @application(inputs=['h_dec'], outputs=['center_y', 'center_x', 'deltaY', 'deltaX'])
    def apply_l(self, h_dec):
        l = self.readout.apply(h_dec)

        center_y, center_x, deltaY, deltaX = self.zoomer.nn2att_wn(l)

        return center_y, center_x, deltaY, deltaX
开发者ID:ablavatski,项目名称:draw,代码行数:45,代码来源:model.py

示例4: main

# 需要导入模块: from attention import ZoomableAttentionWindow [as 别名]
# 或者: from attention.ZoomableAttentionWindow import read [as 别名]
def main(name, epochs, batch_size, learning_rate):
    if name is None:
        name = "att-rw" 

    print("\nRunning experiment %s" % name)
    print("         learning rate: %5.3f" % learning_rate) 
    print()


    #------------------------------------------------------------------------

    img_height, img_width = 28, 28
    
    read_N = 12
    write_N = 14

    inits = {
        #'weights_init': Orthogonal(),
        'weights_init': IsotropicGaussian(0.001),
        'biases_init': Constant(0.),
    }
    
    x_dim = img_height * img_width

    reader = ZoomableAttentionWindow(img_height, img_width,  read_N)
    writer = ZoomableAttentionWindow(img_height, img_width, write_N)

    # Parameterize the attention reader and writer
    mlpr = MLP(activations=[Tanh(), Identity()], 
                dims=[x_dim, 50, 5], 
                name="RMLP",
                **inits)
    mlpw = MLP(activations=[Tanh(), Identity()],
                dims=[x_dim, 50, 5],
                name="WMLP",
                **inits)

    # MLP between the reader and writer
    mlp = MLP(activations=[Tanh(), Identity()],
                dims=[read_N**2, 300, write_N**2],
                name="MLP",
                **inits)

    for brick in [mlpr, mlpw, mlp]:
        brick.allocate()
        brick.initialize()

    #------------------------------------------------------------------------
    x = tensor.matrix('features')

    hr = mlpr.apply(x)
    hw = mlpw.apply(x)

    center_y, center_x, delta, sigma, gamma = reader.nn2att(hr)
    r = reader.read(x, center_y, center_x, delta, sigma)

    h = mlp.apply(r)

    center_y, center_x, delta, sigma, gamma = writer.nn2att(hw)
    c = writer.write(h, center_y, center_x, delta, sigma) / gamma
    x_recons = T.nnet.sigmoid(c)

    cost = BinaryCrossEntropy().apply(x, x_recons)
    cost.name = "cost"

    #------------------------------------------------------------
    cg = ComputationGraph([cost])
    params = VariableFilter(roles=[PARAMETER])(cg.variables)

    algorithm = GradientDescent(
        cost=cost, 
        params=params,
        step_rule=CompositeRule([
            RemoveNotFinite(),
            Adam(learning_rate),
            StepClipping(3.), 
        ])
        #step_rule=RMSProp(learning_rate),
        #step_rule=Momentum(learning_rate=learning_rate, momentum=0.95)
    )


    #------------------------------------------------------------------------
    # Setup monitors
    monitors = [cost]
    #for v in [center_y, center_x, log_delta, log_sigma, log_gamma]:
    #    v_mean = v.mean()
    #    v_mean.name = v.name
    #    monitors += [v_mean]
    #    monitors += [aggregation.mean(v)]

    train_monitors = monitors[:]
    train_monitors += [aggregation.mean(algorithm.total_gradient_norm)]
    train_monitors += [aggregation.mean(algorithm.total_step_norm)]

    # Live plotting...
    plot_channels = [
        ["cost"],
    ]

#.........这里部分代码省略.........
开发者ID:Philip-Bachman,项目名称:NN-Python,代码行数:103,代码来源:run-att-rw.py


注:本文中的attention.ZoomableAttentionWindow.read方法示例由纯净天空整理自Github/MSDocs等开源代码及文档管理平台,相关代码片段筛选自各路编程大神贡献的开源项目,源码版权归原作者所有,传播和使用请参考对应项目的License;未经允许,请勿转载。