Python tensorflow.float32() Examples

def stackedRNN(self, x, dropout, scope, embedding_size, sequence_length, hidden_units):
        n_hidden=hidden_units
        n_layers=3
        # Prepare data shape to match `static_rnn` function requirements
        x = tf.unstack(tf.transpose(x, perm=[1, 0, 2]))
        # print(x)
        # Define lstm cells with tensorflow
        # Forward direction cell

        with tf.name_scope("fw"+scope),tf.variable_scope("fw"+scope):
            stacked_rnn_fw = []
            for _ in range(n_layers):
                fw_cell = tf.nn.rnn_cell.BasicLSTMCell(n_hidden, forget_bias=1.0, state_is_tuple=True)
                lstm_fw_cell = tf.contrib.rnn.DropoutWrapper(fw_cell,output_keep_prob=dropout)
                stacked_rnn_fw.append(lstm_fw_cell)
            lstm_fw_cell_m = tf.nn.rnn_cell.MultiRNNCell(cells=stacked_rnn_fw, state_is_tuple=True)

            outputs, _ = tf.nn.static_rnn(lstm_fw_cell_m, x, dtype=tf.float32)
        return outputs[-1] 

def wrap_variable(self, var):
        """wrap layer.w into variables"""
        val = self.lay.w.get(var, None)
        if val is None:
            shape = self.lay.wshape[var]
            args = [0., 1e-2, shape]
            if 'moving_mean' in var:
                val = np.zeros(shape)
            elif 'moving_variance' in var:
                val = np.ones(shape)
            else:
                val = np.random.normal(*args)
            self.lay.w[var] = val.astype(np.float32)
            self.act = 'Init '
        if not self.var: return

        val = self.lay.w[var]
        self.lay.w[var] = tf.constant_initializer(val)
        if var in self._SLIM: return
        with tf.variable_scope(self.scope):
            self.lay.w[var] = tf.get_variable(var,
                shape = self.lay.wshape[var],
                dtype = tf.float32,
                initializer = self.lay.w[var]) 

def call(self, x):
        if (self.size == None) or (self.mode == 'sum'):
            self.size = int(x.shape[-1])

        position_j = 1. / \
            K.pow(10000., 2 * K.arange(self.size / 2, dtype='float32') / self.size)
        position_j = K.expand_dims(position_j, 0)

        position_i = tf.cumsum(K.ones_like(x[:, :, 0]), 1) - 1
        position_i = K.expand_dims(position_i, 2)
        position_ij = K.dot(position_i, position_j)
        outputs = K.concatenate(
            [K.cos(position_ij), K.sin(position_ij)], 2)

        if self.mode == 'sum':
            if self.scale:
                outputs = outputs * outputs ** 0.5
            return x + outputs
        elif self.mode == 'concat':
            return K.concatenate([outputs, x], 2) 

def _build_input(self):
        self.tails = tf.placeholder(tf.int32, [None])
        self.heads = tf.placeholder(tf.int32, [None])
        self.targets = tf.one_hot(indices=self.heads, depth=self.num_entity)
            
        if not self.query_is_language:
            self.queries = tf.placeholder(tf.int32, [None, self.num_step])
            self.query_embedding_params = tf.Variable(self._random_uniform_unit(
                                                          self.num_query + 1, # <END> token 
                                                          self.query_embed_size), 
                                                      dtype=tf.float32)
        
            rnn_inputs = tf.nn.embedding_lookup(self.query_embedding_params, 
                                                self.queries)
        else:
            self.queries = tf.placeholder(tf.int32, [None, self.num_step, self.num_word])
            self.vocab_embedding_params = tf.Variable(self._random_uniform_unit(
                                                          self.num_vocab + 1, # <END> token
                                                          self.vocab_embed_size),
                                                      dtype=tf.float32)
            embedded_query = tf.nn.embedding_lookup(self.vocab_embedding_params, 
                                                    self.queries)
            rnn_inputs = tf.reduce_mean(embedded_query, axis=2)

        return rnn_inputs 

def noise_input_fn(params):
    """Input function for generating samples for PREDICT mode.

  Generates a single Tensor of fixed random noise. Use tf.data.Dataset to
  signal to the estimator when to terminate the generator returned by
  predict().

  Args:
    params: param `dict` passed by TPUEstimator.

  Returns:
    1-element `dict` containing the randomly generated noise.
  """

    # random noise
    np.random.seed(0)
    noise_dataset = tf.data.Dataset.from_tensors(tf.constant(
        np.random.randn(params['batch_size'], FLAGS.noise_dim), dtype=tf.float32))
    noise = noise_dataset.make_one_shot_iterator().get_next()
    return {'random_noise': noise}, None 

def _mapper(example_proto):
  features = {
      'samples': tf.FixedLenSequenceFeature([1], tf.float32, allow_missing=True),
      'label': tf.FixedLenSequenceFeature([], tf.string, allow_missing=True)
  }
  example = tf.parse_single_example(example_proto, features)

  wav = example['samples'][:, 0]

  wav = wav[:16384]
  wav_len = tf.shape(wav)[0]
  wav = tf.pad(wav, [[0, 16384 - wav_len]])

  label = tf.reduce_join(example['label'], 0)

  return wav, label 

def autosummary(name, value):
    id = name.replace('/', '_')
    if is_tf_expression(value):
        with tf.name_scope('summary_' + id), tf.device(value.device):
            update_op = _create_autosummary_var(name, value)
            with tf.control_dependencies([update_op]):
                return tf.identity(value)
    else: # python scalar or numpy array
        if name not in _autosummary_immediate:
            with absolute_name_scope('Autosummary/' + id), tf.device(None), tf.control_dependencies(None):
                update_value = tf.placeholder(tf.float32)
                update_op = _create_autosummary_var(name, update_value)
                _autosummary_immediate[name] = update_op, update_value
        update_op, update_value = _autosummary_immediate[name]
        run(update_op, {update_value: np.float32(value)})
        return value

# Create the necessary ops to include autosummaries in TensorBoard report.
# Note: This should be done only once per graph. 

def _create_autosummary_var(name, value_expr):
    assert not _autosummary_finalized
    v = tf.cast(value_expr, tf.float32)
    if v.shape.ndims is 0:
        v = [v, np.float32(1.0)]
    elif v.shape.ndims is 1:
        v = [tf.reduce_sum(v), tf.cast(tf.shape(v)[0], tf.float32)]
    else:
        v = [tf.reduce_sum(v), tf.reduce_prod(tf.cast(tf.shape(v), tf.float32))]
    v = tf.cond(tf.is_finite(v[0]), lambda: tf.stack(v), lambda: tf.zeros(2))
    with tf.control_dependencies(None):
        var = tf.Variable(tf.zeros(2)) # [numerator, denominator]
    update_op = tf.cond(tf.is_variable_initialized(var), lambda: tf.assign_add(var, v), lambda: tf.assign(var, v))
    if name in _autosummary_vars:
        _autosummary_vars[name].append(var)
    else:
        _autosummary_vars[name] = [var]
    return update_op

#----------------------------------------------------------------------------
# Call filewriter.add_summary() with all summaries in the default graph,
# automatically finalizing and merging them on the first call. 

def minibatch_stddev_layer(x, group_size=4):
    with tf.variable_scope('MinibatchStddev'):
        group_size = tf.minimum(group_size, tf.shape(x)[0])     # Minibatch must be divisible by (or smaller than) group_size.
        s = x.shape                                             # [NCHW]  Input shape.
        y = tf.reshape(x, [group_size, -1, s[1], s[2], s[3]])   # [GMCHW] Split minibatch into M groups of size G.
        y = tf.cast(y, tf.float32)                              # [GMCHW] Cast to FP32.
        y -= tf.reduce_mean(y, axis=0, keep_dims=True)           # [GMCHW] Subtract mean over group.
        y = tf.reduce_mean(tf.square(y), axis=0)                # [MCHW]  Calc variance over group.
        y = tf.sqrt(y + 1e-8)                                   # [MCHW]  Calc stddev over group.
        y = tf.reduce_mean(y, axis=[1,2,3], keep_dims=True)      # [M111]  Take average over fmaps and pixels.
        y = tf.cast(y, x.dtype)                                 # [M111]  Cast back to original data type.
        y = tf.tile(y, [group_size, 1, s[2], s[3]])             # [N1HW]  Replicate over group and pixels.
        return tf.concat([x, y], axis=1)                        # [NCHW]  Append as new fmap.

#----------------------------------------------------------------------------
# Generator network used in the paper. 

def __init__(self, resolution=1024, num_channels=3, dtype='uint8', dynamic_range=[0,255], label_size=0, label_dtype='float32'):
        self.resolution         = resolution
        self.resolution_log2    = int(np.log2(resolution))
        self.shape              = [num_channels, resolution, resolution]
        self.dtype              = dtype
        self.dynamic_range      = dynamic_range
        self.label_size         = label_size
        self.label_dtype        = label_dtype
        self._tf_minibatch_var  = None
        self._tf_lod_var        = None
        self._tf_minibatch_np   = None
        self._tf_labels_np      = None

        assert self.resolution == 2 ** self.resolution_log2
        with tf.name_scope('Dataset'):
            self._tf_minibatch_var = tf.Variable(np.int32(0), name='minibatch_var')
            self._tf_lod_var = tf.Variable(np.int32(0), name='lod_var') 

def build_pgd_attack(self, eps):
        victim_embeddings = tf.constant(self.victim_embeddings, dtype=tf.float32)

        def one_step_attack(image, grad):
            """
            core components of this attack are:
            (a) PGD adversarial attack (https://arxiv.org/pdf/1706.06083.pdf)
            (b) momentum (https://arxiv.org/pdf/1710.06081.pdf)
            (c) input diversity (https://arxiv.org/pdf/1803.06978.pdf)
            """
            orig_image = image
            image = self.structure(image)
            image = (image - 127.5) / 128.0
            image = image + tf.random_uniform(tf.shape(image), minval=-1e-2, maxval=1e-2)
            prelogits, _ = self.network.inference(image, 1.0, False, bottleneck_layer_size=512)
            embeddings = tf.nn.l2_normalize(prelogits, 1, 1e-10, name='embeddings')

            embeddings = tf.reshape(embeddings[0], [512, 1])
            objective = tf.reduce_mean(tf.matmul(victim_embeddings, embeddings))  # to be maximized

            noise, = tf.gradients(objective, orig_image)

            noise = noise / tf.reduce_mean(tf.abs(noise), [1, 2, 3], keep_dims=True)
            noise = 0.9 * grad + noise

            adv = tf.clip_by_value(orig_image + tf.sign(noise) * 1.0, lower_bound, upper_bound)
            return adv, noise

        input = tf.to_float(self.image_batch)
        lower_bound = tf.clip_by_value(input - eps, 0, 255.)
        upper_bound = tf.clip_by_value(input + eps, 0, 255.)

        with tf.variable_scope(tf.get_variable_scope(), reuse=tf.AUTO_REUSE):
            adv, _ = tf.while_loop(
                lambda _, __: True, one_step_attack,
                (input, tf.zeros_like(input)),
                back_prop=False,
                maximum_iterations=100,
                parallel_iterations=1)
        self.adv_image = adv
        return adv 

def test_generate_np_caches_graph_computation_for_eps_clip_or_xi(self):

        x_val = np.random.rand(1, 2)
        x_val = np.array(x_val, dtype=np.float32)

        self.attack.generate_np(x_val, eps=.3, num_iterations=10,
                                clip_max=-5.0, clip_min=-5.0,
                                xi=1e-6)

        old_grads = tf.gradients

        def fn(*x, **y):
            raise RuntimeError()
        tf.gradients = fn

        self.attack.generate_np(x_val, eps=.2, num_iterations=10,
                                clip_max=-4.0, clip_min=-4.0,
                                xi=1e-5)

        tf.gradients = old_grads 

def test_attack_strength(self):
        """
        If clipping is not done at each iteration (not passing clip_min and
        clip_max to fgm), this attack fails by
        np.mean(orig_labels == new_labels) == .39.
        """
        x_val = np.random.rand(100, 2)
        x_val = np.array(x_val, dtype=np.float32)

        x_adv = self.attack.generate_np(x_val, eps=1.0, ord=np.inf,
                                        clip_min=0.5, clip_max=0.7,
                                        nb_iter=5)

        orig_labs = np.argmax(self.sess.run(self.model(x_val)), axis=1)
        new_labs = np.argmax(self.sess.run(self.model(x_adv)), axis=1)
        self.assertTrue(np.mean(orig_labs == new_labs) < 0.1) 

def test_generate_np_targeted_gives_adversarial_example(self):
        x_val = np.random.rand(100, 2)
        x_val = np.array(x_val, dtype=np.float32)

        feed_labs = np.zeros((100, 2))
        feed_labs[np.arange(100), np.random.randint(0, 1, 100)] = 1
        x_adv = self.attack.generate_np(x_val, max_iterations=100,
                                        binary_search_steps=3,
                                        initial_const=1,
                                        clip_min=-5, clip_max=5,
                                        batch_size=100, y_target=feed_labs)

        new_labs = np.argmax(self.sess.run(self.model(x_adv)), axis=1)

        self.assertTrue(np.mean(np.argmax(feed_labs, axis=1) == new_labs)
                        > 0.9) 

def test_generate_gives_adversarial_example(self):

        x_val = np.random.rand(100, 2)
        x_val = np.array(x_val, dtype=np.float32)

        orig_labs = np.argmax(self.sess.run(self.model(x_val)), axis=1)
        feed_labs = np.zeros((100, 2))
        feed_labs[np.arange(100), orig_labs] = 1
        x = tf.placeholder(tf.float32, x_val.shape)
        y = tf.placeholder(tf.float32, feed_labs.shape)

        x_adv_p = self.attack.generate(x, max_iterations=100,
                                       binary_search_steps=3,
                                       initial_const=1,
                                       clip_min=-5, clip_max=5,
                                       batch_size=100, y=y)
        self.assertEqual(x_val.shape, x_adv_p.shape)
        x_adv = self.sess.run(x_adv_p, {x: x_val, y: feed_labs})

        new_labs = np.argmax(self.sess.run(self.model(x_adv)), axis=1)

        self.assertTrue(np.mean(orig_labs == new_labs) < 0.1) 

def test_generate_np_targeted_gives_adversarial_example(self):
        x_val = np.random.rand(100, 2)
        x_val = np.array(x_val, dtype=np.float32)

        feed_labs = np.zeros((100, 2))
        feed_labs[np.arange(100), np.random.randint(0, 1, 100)] = 1
        x_adv = self.attack.generate_np(x_val, max_iterations=100,
                                        binary_search_steps=3,
                                        initial_const=1,
                                        clip_min=-5, clip_max=5,
                                        batch_size=100, y_target=feed_labs)

        new_labs = np.argmax(self.sess.run(self.model(x_adv)), axis=1)

        self.assertTrue(np.mean(np.argmax(feed_labs, axis=1) == new_labs) >
                        0.9) 

def test_generate_gives_adversarial_example(self):

        x_val = np.random.rand(100, 2)
        x_val = np.array(x_val, dtype=np.float32)

        orig_labs = np.argmax(self.sess.run(self.model(x_val)), axis=1)
        feed_labs = np.zeros((100, 2))
        feed_labs[np.arange(100), orig_labs] = 1
        x = tf.placeholder(tf.float32, x_val.shape)
        y = tf.placeholder(tf.float32, feed_labs.shape)

        x_adv_p = self.attack.generate(x, max_iterations=100,
                                       binary_search_steps=3,
                                       initial_const=1,
                                       clip_min=-5, clip_max=5,
                                       batch_size=100, y=y)
        self.assertEqual(x_val.shape, x_adv_p.shape)
        x_adv = self.sess.run(x_adv_p, {x: x_val, y: feed_labs})

        new_labs = np.argmax(self.sess.run(self.model(x_adv)), axis=1)

        self.assertTrue(np.mean(orig_labs == new_labs) < 0.1) 

def test_generate_gives_adversarial_example(self):

        x_val = np.random.rand(100, 2)
        x_val = np.array(x_val, dtype=np.float32)

        orig_labs = np.argmax(self.sess.run(self.model(x_val)), axis=1)
        x = tf.placeholder(tf.float32, x_val.shape)

        x_adv_p = self.attack.generate(x, over_shoot=0.02, max_iter=50,
                                       nb_candidate=2, clip_min=-5, clip_max=5)
        self.assertEqual(x_val.shape, x_adv_p.shape)
        x_adv = self.sess.run(x_adv_p, {x: x_val})

        new_labs = np.argmax(self.sess.run(self.model(x_adv)), axis=1)

        self.assertTrue(np.mean(orig_labs == new_labs) < 0.1) 

def test_attack_strength(self):
        """
        If clipping is not done at each iteration (not using clip_min and
        clip_max), this attack fails by
        np.mean(orig_labels == new_labels) == .5
        """
        x_val = np.random.rand(100, 2)
        x_val = np.array(x_val, dtype=np.float32)

        x_adv = self.attack.generate_np(x_val, eps=1.0, eps_iter=0.05,
                                        clip_min=0.5, clip_max=0.7,
                                        nb_iter=5)

        orig_labs = np.argmax(self.sess.run(self.model(x_val)), axis=1)
        new_labs = np.argmax(self.sess.run(self.model(x_adv)), axis=1)
        self.assertTrue(np.mean(orig_labs == new_labs) < 0.1) 

def test_generate_np_targeted_gives_adversarial_example(self):
        x_val = np.random.rand(100, 2)
        x_val = np.array(x_val, dtype=np.float32)

        feed_labs = np.zeros((100, 2))
        feed_labs[np.arange(100), np.random.randint(0, 1, 100)] = 1
        x_adv = self.attack.generate_np(x_val, max_iterations=100,
                                        binary_search_steps=3,
                                        initial_const=1,
                                        clip_min=-5, clip_max=5,
                                        batch_size=100, y_target=feed_labs)

        new_labs = np.argmax(self.sess.run(self.model(x_adv)), axis=1)

        self.assertTrue(np.mean(np.argmax(feed_labs, axis=1) == new_labs)
                        > 0.9) 

def test_generate_targeted_gives_adversarial_example(self):
        x_val = np.random.rand(100, 2)
        x_val = np.array(x_val, dtype=np.float32)

        feed_labs = np.zeros((100, 2))
        feed_labs[np.arange(100), np.random.randint(0, 1, 100)] = 1
        x = tf.placeholder(tf.float32, x_val.shape)
        y = tf.placeholder(tf.float32, feed_labs.shape)

        x_adv_p = self.attack.generate(x, max_iterations=100,
                                       binary_search_steps=3,
                                       initial_const=1,
                                       clip_min=-5, clip_max=5,
                                       batch_size=100, y_target=y)
        self.assertEqual(x_val.shape, x_adv_p.shape)
        x_adv = self.sess.run(x_adv_p, {x: x_val, y: feed_labs})

        new_labs = np.argmax(self.sess.run(self.model(x_adv)), axis=1)

        self.assertTrue(np.mean(np.argmax(feed_labs, axis=1) == new_labs)
                        > 0.9) 

def set_input_shape(self, input_shape):
        batch_size, rows, cols, input_channels = input_shape
        kernel_shape = tuple(self.kernel_shape) + (input_channels,
                                                   self.output_channels)
        assert len(kernel_shape) == 4
        assert all(isinstance(e, int) for e in kernel_shape), kernel_shape
        init = tf.random_normal(kernel_shape, dtype=tf.float32)
        init = init / tf.sqrt(1e-7 + tf.reduce_sum(tf.square(init),
                                                   axis=(0, 1, 2)))
        self.kernels = tf.Variable(init)
        self.b = tf.Variable(
            np.zeros((self.output_channels,)).astype('float32'))
        input_shape = list(input_shape)
        input_shape[0] = 1
        dummy_batch = tf.zeros(input_shape)
        dummy_output = self.fprop(dummy_batch)
        output_shape = [int(e) for e in dummy_output.get_shape()]
        output_shape[0] = batch_size
        self.output_shape = tuple(output_shape) 

def compute_mfcc(audio, **kwargs):
    """
    Compute the MFCC for a given audio waveform. This is
    identical to how DeepSpeech does it, but does it all in
    TensorFlow so that we can differentiate through it.
    """

    batch_size, size = audio.get_shape().as_list()
    audio = tf.cast(audio, tf.float32)

    # 1. Pre-emphasizer, a high-pass filter
    audio = tf.concat((audio[:, :1], audio[:, 1:] - 0.97*audio[:, :-1], np.zeros((batch_size,1000),dtype=np.float32)), 1)

    # 2. windowing into frames of 320 samples, overlapping
    windowed = tf.stack([audio[:, i:i+400] for i in range(0,size-320,160)],1)

    # 3. Take the FFT to convert to frequency space
    ffted = tf.spectral.rfft(windowed, [512])
    ffted = 1.0 / 512 * tf.square(tf.abs(ffted))

    # 4. Compute the Mel windowing of the FFT
    energy = tf.reduce_sum(ffted,axis=2)+1e-30
    filters = np.load("filterbanks.npy").T
    feat = tf.matmul(ffted, np.array([filters]*batch_size,dtype=np.float32))+1e-30

    # 5. Take the DCT again, because why not
    feat = tf.log(feat)
    feat = tf.spectral.dct(feat, type=2, norm='ortho')[:,:,:26]

    # 6. Amplify high frequencies for some reason
    _,nframes,ncoeff = feat.get_shape().as_list()
    n = np.arange(ncoeff)
    lift = 1 + (22/2.)*np.sin(np.pi*n/22)
    feat = lift*feat
    width = feat.get_shape().as_list()[1]

    # 7. And now stick the energy next to the features
    feat = tf.concat((tf.reshape(tf.log(energy),(-1,width,1)), feat[:, :, 1:]), axis=2)
    
    return feat 

def get_logits(new_input, length, first=[]):
    """
    Compute the logits for a given waveform.

    First, preprocess with the TF version of MFC above,
    and then call DeepSpeech on the features.
    """
    # new_input = tf.Print(new_input, [tf.shape(new_input)])

    # We need to init DeepSpeech the first time we're called
    if first == []:
        first.append(False)
        # Okay, so this is ugly again.
        # We just want it to not crash.
        tf.app.flags.FLAGS.alphabet_config_path = "DeepSpeech/data/alphabet.txt"
        DeepSpeech.initialize_globals()
        print('initialized deepspeech globals')

    batch_size = new_input.get_shape()[0]

    # 1. Compute the MFCCs for the input audio
    # (this is differentable with our implementation above)
    empty_context = np.zeros((batch_size, 9, 26), dtype=np.float32)
    new_input_to_mfcc = compute_mfcc(new_input)[:, ::2]
    features = tf.concat((empty_context, new_input_to_mfcc, empty_context), 1)

    # 2. We get to see 9 frames at a time to make our decision,
    # so concatenate them together.
    features = tf.reshape(features, [new_input.get_shape()[0], -1])
    features = tf.stack([features[:, i:i+19*26] for i in range(0,features.shape[1]-19*26+1,26)],1)
    features = tf.reshape(features, [batch_size, -1, 19*26])

    # 3. Whiten the data
    mean, var = tf.nn.moments(features, axes=[0,1,2])
    features = (features-mean)/(var**.5)

    # 4. Finally we process it with DeepSpeech
    logits = DeepSpeech.BiRNN(features, length, [0]*10)

    return logits 

def create_attention_mask_from_input_mask(from_tensor, to_mask):
    """Create 3D attention mask from a 2D tensor mask.

    Args:
      from_tensor: 2D or 3D Tensor of shape [batch_size, from_seq_length, ...].
      to_mask: int32 Tensor of shape [batch_size, to_seq_length].

    Returns:
      float Tensor of shape [batch_size, from_seq_length, to_seq_length].
    """
    from_shape = get_shape_list(from_tensor, expected_rank=[2, 3])
    batch_size = from_shape[0]
    from_seq_length = from_shape[1]

    to_shape = get_shape_list(to_mask, expected_rank=2)
    to_seq_length = to_shape[1]

    to_mask = tf.cast(
        tf.reshape(to_mask, [batch_size, 1, to_seq_length]), tf.float32)

    # We don't assume that `from_tensor` is a mask (although it could be). We
    # don't actually care if we attend *from* padding tokens (only *to* padding)
    # tokens so we create a tensor of all ones.
    #
    # `broadcast_ones` = [batch_size, from_seq_length, 1]
    broadcast_ones = tf.ones(
        shape=[batch_size, from_seq_length, 1], dtype=tf.float32)

    # Here we broadcast along two dimensions to create the mask.
    mask = broadcast_ones * to_mask

    return mask 

def __init__(
        self, sequence_length, vocab_size, embedding_size, hidden_units, l2_reg_lambda, batch_size, trainableEmbeddings):

        # Placeholders for input, output and dropout
        self.input_x1 = tf.placeholder(tf.int32, [None, sequence_length], name="input_x1")
        self.input_x2 = tf.placeholder(tf.int32, [None, sequence_length], name="input_x2")
        self.input_y = tf.placeholder(tf.float32, [None], name="input_y")
        self.dropout_keep_prob = tf.placeholder(tf.float32, name="dropout_keep_prob")

        # Keeping track of l2 regularization loss (optional)
        l2_loss = tf.constant(0.0, name="l2_loss")
          
        # Embedding layer
        with tf.name_scope("embedding"):
            self.W = tf.Variable(
                tf.constant(0.0, shape=[vocab_size, embedding_size]),
                trainable=trainableEmbeddings,name="W")
            self.embedded_words1 = tf.nn.embedding_lookup(self.W, self.input_x1)
            self.embedded_words2 = tf.nn.embedding_lookup(self.W, self.input_x2)
        print self.embedded_words1
        # Create a convolution + maxpool layer for each filter size
        with tf.name_scope("output"):
            self.out1=self.stackedRNN(self.embedded_words1, self.dropout_keep_prob, "side1", embedding_size, sequence_length, hidden_units)
            self.out2=self.stackedRNN(self.embedded_words2, self.dropout_keep_prob, "side2", embedding_size, sequence_length, hidden_units)
            self.distance = tf.sqrt(tf.reduce_sum(tf.square(tf.subtract(self.out1,self.out2)),1,keep_dims=True))
            self.distance = tf.div(self.distance, tf.add(tf.sqrt(tf.reduce_sum(tf.square(self.out1),1,keep_dims=True)),tf.sqrt(tf.reduce_sum(tf.square(self.out2),1,keep_dims=True))))
            self.distance = tf.reshape(self.distance, [-1], name="distance")
        with tf.name_scope("loss"):
            self.loss = self.contrastive_loss(self.input_y,self.distance, batch_size)
        #### Accuracy computation is outside of this class.
        with tf.name_scope("accuracy"):
            self.temp_sim = tf.subtract(tf.ones_like(self.distance),tf.rint(self.distance), name="temp_sim") #auto threshold 0.5
            correct_predictions = tf.equal(self.temp_sim, self.input_y)
            self.accuracy=tf.reduce_mean(tf.cast(correct_predictions, "float"), name="accuracy") 

def BiRNN(self, x, dropout, scope, embedding_size, sequence_length, hidden_units):
        n_hidden=hidden_units
        n_layers=3
        # Prepare data shape to match `static_rnn` function requirements
        x = tf.unstack(tf.transpose(x, perm=[1, 0, 2]))
        print(x)
        # Define lstm cells with tensorflow
        # Forward direction cell
        with tf.name_scope("fw"+scope),tf.variable_scope("fw"+scope):
            stacked_rnn_fw = []
            for _ in range(n_layers):
                fw_cell = tf.nn.rnn_cell.BasicLSTMCell(n_hidden, forget_bias=1.0, state_is_tuple=True)
                lstm_fw_cell = tf.contrib.rnn.DropoutWrapper(fw_cell,output_keep_prob=dropout)
                stacked_rnn_fw.append(lstm_fw_cell)
            lstm_fw_cell_m = tf.nn.rnn_cell.MultiRNNCell(cells=stacked_rnn_fw, state_is_tuple=True)

        with tf.name_scope("bw"+scope),tf.variable_scope("bw"+scope):
            stacked_rnn_bw = []
            for _ in range(n_layers):
                bw_cell = tf.nn.rnn_cell.BasicLSTMCell(n_hidden, forget_bias=1.0, state_is_tuple=True)
                lstm_bw_cell = tf.contrib.rnn.DropoutWrapper(bw_cell,output_keep_prob=dropout)
                stacked_rnn_bw.append(lstm_bw_cell)
            lstm_bw_cell_m = tf.nn.rnn_cell.MultiRNNCell(cells=stacked_rnn_bw, state_is_tuple=True)
        # Get lstm cell output

        with tf.name_scope("bw"+scope),tf.variable_scope("bw"+scope):
            outputs, _, _ = tf.nn.static_bidirectional_rnn(lstm_fw_cell_m, lstm_bw_cell_m, x, dtype=tf.float32)
        return outputs[-1] 

def __init__(
        self, sequence_length, vocab_size, embedding_size, hidden_units, l2_reg_lambda, batch_size):

        # Placeholders for input, output and dropout
        self.input_x1 = tf.placeholder(tf.int32, [None, sequence_length], name="input_x1")
        self.input_x2 = tf.placeholder(tf.int32, [None, sequence_length], name="input_x2")
        self.input_y = tf.placeholder(tf.float32, [None], name="input_y")
        self.dropout_keep_prob = tf.placeholder(tf.float32, name="dropout_keep_prob")

        # Keeping track of l2 regularization loss (optional)
        l2_loss = tf.constant(0.0, name="l2_loss")
          
        # Embedding layer
        with tf.name_scope("embedding"):
            self.W = tf.Variable(
                tf.random_uniform([vocab_size, embedding_size], -1.0, 1.0),
                trainable=True,name="W")
            self.embedded_chars1 = tf.nn.embedding_lookup(self.W, self.input_x1)
            #self.embedded_chars_expanded1 = tf.expand_dims(self.embedded_chars1, -1)
            self.embedded_chars2 = tf.nn.embedding_lookup(self.W, self.input_x2)
            #self.embedded_chars_expanded2 = tf.expand_dims(self.embedded_chars2, -1)

        # Create a convolution + maxpool layer for each filter size
        with tf.name_scope("output"):
            self.out1=self.BiRNN(self.embedded_chars1, self.dropout_keep_prob, "side1", embedding_size, sequence_length, hidden_units)
            self.out2=self.BiRNN(self.embedded_chars2, self.dropout_keep_prob, "side2", embedding_size, sequence_length, hidden_units)
            self.distance = tf.sqrt(tf.reduce_sum(tf.square(tf.subtract(self.out1,self.out2)),1,keep_dims=True))
            self.distance = tf.div(self.distance, tf.add(tf.sqrt(tf.reduce_sum(tf.square(self.out1),1,keep_dims=True)),tf.sqrt(tf.reduce_sum(tf.square(self.out2),1,keep_dims=True))))
            self.distance = tf.reshape(self.distance, [-1], name="distance")
        with tf.name_scope("loss"):
            self.loss = self.contrastive_loss(self.input_y,self.distance, batch_size)
        #### Accuracy computation is outside of this class.
        with tf.name_scope("accuracy"):
            self.temp_sim = tf.subtract(tf.ones_like(self.distance),tf.rint(self.distance), name="temp_sim") #auto threshold 0.5
            correct_predictions = tf.equal(self.temp_sim, self.input_y)
            self.accuracy=tf.reduce_mean(tf.cast(correct_predictions, "float"), name="accuracy") 

def build(self, input_shape):
        self.bn = tf.keras.layers.BatchNormalization(
            axis=self.axis, epsilon=self.epsilon, center=False, scale=False)
        self.alphas = self.add_weight(shape=(input_shape[-1],), initializer=Zeros(
        ), dtype=tf.float32, name=self.name+'dice_alpha')  # name='alpha_'+self.name
        super(Dice, self).build(input_shape)  # Be sure to call this somewhere! 

def call(self, seq_value_len_list, mask=None, **kwargs):
        if self.supports_masking:
            if mask is None:
                raise ValueError(
                    "When supports_masking=True,input must support masking")
            uiseq_embed_list = seq_value_len_list
            mask = tf.to_float(mask)
            user_behavior_length = tf.reduce_sum(mask, axis=-1, keep_dims=True)
            mask = tf.expand_dims(mask, axis=2)
        else:
            uiseq_embed_list, user_behavior_length = seq_value_len_list

            mask = tf.sequence_mask(user_behavior_length,
                                    self.seq_len_max, dtype=tf.float32)
            mask = tf.transpose(mask, (0, 2, 1))

        embedding_size = uiseq_embed_list.shape[-1]

        mask = tf.tile(mask, [1, 1, embedding_size])

        uiseq_embed_list *= mask
        hist = uiseq_embed_list
        if self.mode == "max":
            return tf.reduce_max(hist, 1, keep_dims=True)

        hist = tf.reduce_sum(hist, 1, keep_dims=False)

        if self.mode == "mean":
            hist = tf.div(hist, user_behavior_length+self.eps)

        hist = tf.expand_dims(hist, axis=1)
        return hist