Python tensorflow.contrib.cudnn_rnn.python.ops.cudnn_rnn_ops.CudnnGRU() Examples

The following are 9 code examples of tensorflow.contrib.cudnn_rnn.python.ops.cudnn_rnn_ops.CudnnGRU(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module tensorflow.contrib.cudnn_rnn.python.ops.cudnn_rnn_ops , or try the search function

Example #1

Source File: test_rnn.py From MMdnn with MIT License

5 votes

def create_symbol(X, num_classes=0, is_training=False, CUDNN=False, 
                  maxf=30000, edim=125, nhid=100, batchs=64):
    word_vectors = tf.contrib.layers.embed_sequence(X, vocab_size=maxf, embed_dim=edim)
    

    word_list = tf.unstack(word_vectors, axis=1)
    
    if not CUDNN:
        cell1 = tf.contrib.rnn.LSTMCell(nhid)
        cell2 = tf.contrib.rnn.GRUCell(nhid)
        stacked_cell = tf.nn.rnn_cell.MultiRNNCell([cell1, cell2])
        outputs, states = tf.nn.static_rnn(stacked_cell, word_list, dtype=tf.float32)
        logits = tf.layers.dense(outputs[-1], 2, activation=None, name='output')
    else:
        # Using cuDNN since vanilla RNN
        from tensorflow.contrib.cudnn_rnn.python.ops import cudnn_rnn_ops
        cudnn_cell = cudnn_rnn_ops.CudnnGRU(num_layers=1, 
                                            num_units=nhid, 
                                            input_size=edim, 
                                            input_mode='linear_input')
        params_size_t = cudnn_cell.params_size()
        params = tf.Variable(tf.random_uniform([params_size_t], -0.1, 0.1), validate_shape=False)   
        input_h = tf.Variable(tf.zeros([1, batchs, nhid]))
        outputs, states = cudnn_cell(input_data=word_list,
                                     input_h=input_h,
                                     params=params)
        logits = tf.layers.dense(outputs[-1], 2, activation=None, name='output')
    
    return logits, logits

Example #2

Source File: cudnn_recurrent.py From GraphicDesignPatternByPython with MIT License

4 votes

def build(self, input_shape):
        super(CuDNNGRU, self).build(input_shape)
        if isinstance(input_shape, list):
            input_shape = input_shape[0]
        input_dim = input_shape[-1]

        from tensorflow.contrib.cudnn_rnn.python.ops import cudnn_rnn_ops
        self._cudnn_gru = cudnn_rnn_ops.CudnnGRU(
            num_layers=1,
            num_units=self.units,
            input_size=input_dim,
            input_mode='linear_input')

        self.kernel = self.add_weight(shape=(input_dim, self.units * 3),
                                      name='kernel',
                                      initializer=self.kernel_initializer,
                                      regularizer=self.kernel_regularizer,
                                      constraint=self.kernel_constraint)
        self.recurrent_kernel = self.add_weight(
            shape=(self.units, self.units * 3),
            name='recurrent_kernel',
            initializer=self.recurrent_initializer,
            regularizer=self.recurrent_regularizer,
            constraint=self.recurrent_constraint)

        self.bias = self.add_weight(shape=(self.units * 6,),
                                    name='bias',
                                    initializer=self.bias_initializer,
                                    regularizer=self.bias_regularizer,
                                    constraint=self.bias_constraint)

        self.kernel_z = self.kernel[:, :self.units]
        self.recurrent_kernel_z = self.recurrent_kernel[:, :self.units]
        self.kernel_r = self.kernel[:, self.units: self.units * 2]
        self.recurrent_kernel_r = self.recurrent_kernel[:,
                                                        self.units:
                                                        self.units * 2]
        self.kernel_h = self.kernel[:, self.units * 2:]
        self.recurrent_kernel_h = self.recurrent_kernel[:, self.units * 2:]

        self.bias_z_i = self.bias[:self.units]
        self.bias_r_i = self.bias[self.units: self.units * 2]
        self.bias_h_i = self.bias[self.units * 2: self.units * 3]
        self.bias_z = self.bias[self.units * 3: self.units * 4]
        self.bias_r = self.bias[self.units * 4: self.units * 5]
        self.bias_h = self.bias[self.units * 5:]

        self.built = True

Example #3