Python Tensorflow：噪声对比估计语言模型

我想将示例中的loss函数更改为

tf.nn.nce\u loss

。查看

tf.nn.nce\u损失
实现：

def nce_loss(weights, biases, inputs, labels, num_sampled, num_classes,
         num_true=1,
         sampled_values=None,
         remove_accidental_hits=False,
         partition_strategy="mod",
         name="nce_loss"):

我想


第三个参数（输入）是语言模型的logits
第四个参数（标签）是语言的下一个单词（self.\u targets）
模型
num_classes是语音大小

但我不知道前两个参数是什么，权重和偏差。我如何使
tf.nn.nce\u loss
适应语言模型？谢谢

########更新
@亚伦：

谢谢，我尝试了以下方法：

loss = tf.reduce_mean(
        tf.nn.nce_loss(softmax_w, softmax_b, logits, tf.reshape(self._targets, [-1,1]),
                                     64, vocab_size))

根据以下文件：


权重：形状的张量[num_classes，dim]或张量列表
物体
其沿维度0的连接具有形状[num_classes，dim]。类嵌入（可能是分区的）

偏差：形状的张量[num_类]。阶级偏见

输入：形状张量[批次大小，尺寸]。正向激活
输入网络的一部分

标签：int64类型的张量和形状[batch\u size，num\u true]。这个
目标类

num_sampled：一个整数。每个类随机抽样的数量
一批

num_classes：一个整数。可能的类数


所以


权重是具有形状（隐藏大小，
人声大小）
偏差为softmax_b，具有形状（声音大小）
输入是logits，它具有形状（批处理大小*num步数，vocab大小）
标签为自。_目标，具有形状（批次大小、数量步数），
因此，我们需要重塑它，tf.重塑（自我目标，[-1,1]）

我的PTBModel模型看起来像

class PTBModel(object):
    def __init__(self, is_training, config):
        self.batch_size = batch_size = config.batch_size
        self.num_steps = num_steps = config.num_steps
        size = config.hidden_size
        vocab_size = config.vocab_size
        self._input_data = tf.placeholder(tf.int32, [batch_size, num_steps])
        self._targets = tf.placeholder(tf.int32, [batch_size, num_steps])

        lstm_cell = rnn_cell.BasicLSTMCell(size, forget_bias=0.0)
        if is_training and config.keep_prob < 1:
            lstm_cell = rnn_cell.DropoutWrapper(lstm_cell, output_keep_prob=config.keep_prob)
        cell = rnn_cell.MultiRNNCell([lstm_cell] * config.num_layers)
        self._initial_state = cell.zero_state(batch_size, tf.float32)
        with tf.device("/cpu:0"):
            embedding = tf.get_variable("embedding", [vocab_size, size])
            inputs = tf.nn.embedding_lookup(embedding, self._input_data)
        if is_training and config.keep_prob < 1:
            inputs = tf.nn.dropout(inputs, config.keep_prob)

        outputs = []
        states = []
        state = self._initial_state
        with tf.variable_scope("RNN"):
            for time_step in range(num_steps):
                if time_step > 0: tf.get_variable_scope().reuse_variables()
                (cell_output, state) = cell(inputs[:, time_step, :], state)
                outputs.append(cell_output)
                states.append(state)
        output = tf.reshape(tf.concat(1, outputs), [-1, size])
        softmax_w = tf.get_variable("softmax_w", [size, vocab_size])
        softmax_b = tf.get_variable("softmax_b", [vocab_size])
        logits = tf.matmul(output, softmax_w) + softmax_b

        '''
        #minimize the average negative log probability using sequence_loss_by_example
        loss = seq2seq.sequence_loss_by_example([logits],
                                                [tf.reshape(self._targets, [-1])],
                                                [tf.ones([batch_size * num_steps])],
                                                vocab_size)

        loss = tf.reduce_mean(
            tf.nn.nce_loss(nce_weights, nce_biases, embed, train_labels,
                                         num_sampled, vocabulary_size))
        weights: A Tensor of shape [num_classes, dim], or a list of Tensor objects 
            whose concatenation along dimension 0 has shape [num_classes, dim]. The (possibly-partitioned) class embeddings.
        biases: A Tensor of shape [num_classes]. The class biases.
        inputs: A Tensor of shape [batch_size, dim]. The forward activations of the input network.
        labels: A Tensor of type int64 and shape [batch_size, num_true]. The target classes.
        num_sampled: An int. The number of classes to randomly sample per batch.
        num_classes: An int. The number of possible classes.

        '''
        loss = tf.reduce_mean(
            tf.nn.nce_loss(softmax_w, softmax_b, logits, tf.reshape(self._targets, [-1,1]),
                                         64, vocab_size))


        self._cost = cost = tf.reduce_sum(loss) / batch_size
        self._final_state = states[-1]
        if not is_training:
            return
        self._lr = tf.Variable(0.0, trainable=False)
        tvars = tf.trainable_variables()
        grads, _ = tf.clip_by_global_norm(tf.gradients(cost, tvars),
                                          config.max_grad_norm)
        optimizer = tf.train.GradientDescentOptimizer(self.lr)
        self._train_op = optimizer.apply_gradients(zip(grads, tvars))

class PTBModel（对象）：
定义初始化（自我、is\U培训、配置）：
self.batch\u size=batch\u size=config.batch\u size
self.num\u steps=num\u steps=config.num\u steps
size=config.hidden\u size
vocab_size=config.vocab_size
self.\u input\u data=tf.placeholder（tf.int32，[batch\u size，num\u steps]）
self.\u targets=tf.placeholder（tf.int32，[batch\u size，num\u steps]）
lstm\u单元=rnn\u单元。基本STMCELL（大小，偏差=0.0）
如果是培训和配置保持问题<1：
lstm\u单元=rnn\u单元.dropoutrapper（lstm\u单元，输出保持保持保持保持保持保持配置）
cell=rnn_cell.MultiRNNCell（[lstm_cell]*config.num_layers）
self.\u初始\u状态=cell.zero\u状态（批处理大小，tf.float32）
使用tf.device（“/cpu:0”）：
嵌入=tf.get_变量（“嵌入”，[vocab_大小，大小]）
inputs=tf.nn.embedding\u lookup（嵌入，自输入数据）
如果是培训和配置保持问题<1：
输入=tf.nn.dropout（输入，配置keep_prob）
输出=[]
国家=[]
状态=自身。初始状态
使用tf.variable_scope（“RNN”）：
对于范围内的时间步长（num步长）：
如果时间步长>0:tf.get_variable_scope（）.reuse_variables（）
（单元格输出，状态）=单元格（输入[：，时间步长，：]，状态）
输出。追加（单元格输出）
states.append（state）
输出=tf.reforme（tf.concat（1，输出），[-1，大小]）
softmax_w=tf.get_变量（“softmax_w，[size，vocab_size]）
softmax_b=tf.get_变量（“softmax_b”，[vocab_size]）
logits=tf.matmul（输出，softmax_w）+softmax_b
'''
#通过示例使用序列损失最小化平均负对数概率
损耗=序列损耗按示例（[logits]，
[tf.重塑（自我目标，[-1]），
[tf.ones（[batch_size*num_steps]），
人声大小）
损失=tf.reduce_平均值(
tf.nn.nce_损失（nce_权重、nce_偏差、嵌入、训练标签、，
数量（抽样，词汇（大小））
权重：形状[num_classes，dim]的张量或张量对象列表
其沿维度0的连接具有形状[num_classes，dim]。类嵌入（可能是分区的）。
偏差：形状的张量[num_类]。阶级偏见。
输入：形状张量[批次大小，尺寸]。输入网络的正向激活。
标签：int64类型的张量和形状[batch\u size，num\u true]。目标类。
num_sampled：一个整数。每批随机抽样的类数。
num_classes：一个整数。可能的类数。
'''
损失=tf.reduce_平均值(
tf.nn.nce_损失（softmax_w，softmax_b，logits，tf.Reformation（自我目标，[-1,1]），
64个，B_尺寸）
自身成本=成本=tf.减少总量（损失）/批量
self.\u final\u state=状态[-1]
如果不是，则是_培训：
返回
self._lr=tf.变量（0.0，可训练=False）
tvars=tf.可训练的_变量（）
梯度，按全球标准（梯度（成本，tVAR）剪裁，
配置（最大梯度标准）
优化器=tf.train.GradientDescentOptimizer（self.lr）
self.\u train\u op=优化器。应用梯度（zip（梯度，TVAR））

然而，我犯了一个错误

Epoch: 1 Learning rate: 1.000
W tensorflow/core/common_runtime/executor.cc:1102] 0x528c980 Compute status: Invalid argument: Index 9971 at offset 0 in Tindices is out of range
     [[Node: model/nce_loss/embedding_lookup = Gather[Tindices=DT_INT64, Tparams=DT_FLOAT, validate_indices=true, _device="/job:localhost/replica:0/task:0/cpu:0"](model/softmax_w/read, model/nce_loss/concat)]]
W tensorflow/core/common_runtime/executor.cc:1102] 0x528c980 Compute status: Invalid argument: Index 9971 at offset 0 in Tindices is out of range
     [[Node: model/nce_loss/embedding_lookup = Gather[Tindices=DT_INT64, Tparams=DT_FLOAT, validate_indices=true, _device="/job:localhost/replica:0/task:0/cpu:0"](model/softmax_w/read, model/nce_loss/concat)]]
     [[Node: _send_model/RNN/concat_19_0 = _Send[T=DT_FLOAT, client_terminated=true, recv_device="/job:localhost/replica:0/task:0/cpu:0", send_device="/job:localhost/replica:0/task:0/cpu:0", send_device_incarnation=1438650956868917036, tensor_name="model/RNN/concat_19:0", _device="/job:localhost/replica:0/task:0/cpu:0"](model/RNN/concat_19)]]
Traceback (most recent call last):
  File "/home/user/works/workspace/python/ptb_word_lm/ptb_word_lm.py", line 235, in <module>
    tf.app.run()
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/default/_app.py", line 30, in run
    sys.exit(main(sys.argv))
  File "/home/user/works/workspace/python/ptb_word_lm/ptb_word_lm.py", line 225, in main
    verbose=True)
  File "/home/user/works/workspace/python/ptb_word_lm/ptb_word_lm.py", line 189, in run_epoch
    m.initial_state: state})
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 315, in run
    return self._run(None, fetches, feed_dict)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 511, in _run
    feed_dict_string)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 564, in _do_run
    target_list)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 586, in _do_call
    e.code)
tensorflow.python.framework.errors.InvalidArgumentError: Index 9971 at offset 0 in Tindices is out of range
     [[Node: model/nce_loss/embedding_lookup = Gather[Tindices=DT_INT64, Tparams=DT_FLOAT, validate_indices=true, _device="/job:localhost/replica:0/task:0/cpu:0"](model/softmax_w/read, model/nce_loss/concat)]]
Caused by op u'model/nce_loss/embedding_lookup', defined at:
  File "/home/user/works/workspace/python/ptb_word_lm/ptb_word_lm.py", line 235, in <module>
    tf.app.run()
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/default/_app.py", line 30, in run
    sys.exit(main(sys.argv))
  File "/home/user/works/workspace/python/ptb_word_lm/ptb_word_lm.py", line 214, in main
    m = PTBModel(is_training=True, config=config)
  File "/home/user/works/workspace/python/ptb_word_lm/ptb_word_lm.py", line 122, in __init__
    64, vocab_size))
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/nn.py", line 798, in nce_loss
    name=name)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/nn.py", line 660, in _compute_sampled_logits
    weights, all_ids, partition_strategy=partition_strategy)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/embedding_ops.py", line 86, in embedding_lookup
    validate_indices=validate_indices)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/gen_array_ops.py", line 447, in gather
    validate_indices=validate_indices, name=name)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/op_def_library.py", line 655, in apply_op
    op_def=op_def)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 2040, in create_op
    original_op=self._default_original_op, op_def=op_def)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 1087, in __init__
    self._traceback = _extract_stack()

Epoch:1学习率：1.000
W tensorflow/core/common_runtime/executor.cc:1102]0x528c980计算状态：无效参数：Tindices中偏移量0处的索引9971超出范围
[[Node:model/nce\u loss/embedding\u lookup=Gather[Tindices=DT\u INT64，Tparams=DT\u FLOAT，validate\u index=true，\u device=“/job:localhost/replica:0/task:0/cpu:0”]（model/softmax w/read，model/nce\u loss/concat）]]
W tensorflow/core/common_runtime/executor.cc:1102]0x528c980计算状态：无效参数：Tindices中偏移量0处的索引9971超出范围
[[Node:model/nce\u loss/embedding\u lookup=Gather[Tindices=DT\u INT64，Tpar