ptb的代码可以详见gitlab上的tensorflow/models/tutorials下，本文只详解他的数据前处理和模型部分。
1.运行
首先说一下他的运行 ，下载数据集：
http://www.fit.vutbr.cz/~imikolov/rnnlm/simple-examples.tgz
解压出ptb.test/train/valid.txt文件，
运行命令行 ：python ptb_word_lm.py --data_path=【文件存放目录】 --model small即可运行成功
2.数据前处理部分
读代码首先要找到主函数，在ptb_word_lm.py文件中，我们看到main函数
def main(_):
if not FLAGS.data_path:
raise ValueError("Must set --data_path to PTB data directory")
gpus = [
x.name for x in device_lib.list_local_devices() if x.device_type == "GPU" ]
if FLAGS.num_gpus > len(gpus):
raise ValueError(
"Your machine has only %d gpus " "which is less than the requested --num_gpus=%d." % (len(gpus), FLAGS.num_gpus))
raw_data = reader.ptb_raw_data(FLAGS.data_path)
train_data, valid_data, test_data, _ = raw_data
显然，与前处理相关的只有ptb_raw_data()函数，在reader文件中，找到相关的四个函数：
import tensorflow as tf
def _read_words(filename):
with tf.gfile.GFile(filename, "r") as f:
return f.read().decode("utf-8").replace("\n", "<eos>").split()
def _build_vocab(filename):
data = _read_words(filename)
counter = collections.Counter(data)
count_pairs = sorted(counter.items(), key=lambda x: (-x[1], x[0]))
words, _ = list(zip(*count_pairs))
word_to_id = dict(zip(words, range(len(words))))
return word_to_id
def _file_to_word_ids(filename, word_to_id):
data = _read_words(filename)
return [word_to_id[word] for word in data if word in word_to_id]
def ptb_raw_data(data_path=None):
train_path = os.path.join(data_path, "ptb.train.txt")
valid_path = os.path.join(data_path, "ptb.valid.txt")
test_path = os.path.join(data_path, "ptb.test.txt")
word_to_id = _build_vocab(train_path)
train_data = _file_to_word_ids(train_path, word_to_id)
valid_data = _file_to_word_ids(valid_path, word_to_id)
test_data = _file_to_word_ids(test_path, word_to_id)
vocabulary = len(word_to_id)
return train_data, valid_data, test_data, vocabulary
_build_vocab函数首先读入文件中所有数据（如：the,a,...），然后以每个单词作为键值，对应给其一个整形数值('the':32021,'a':323)，count_pairs将其转化为对应的元组（'the',32021）,('a',323),建立词表
_file_to_word_ids函数讲文件转化为此表中词所对应的整形数据
返回对应的训练集，验证集和测试集对应的整形数据文件和词表对应长度。
with tf.Graph().as_default():
initializer = tf.random_uniform_initializer(-config.init_scale, config.init_scale)
with tf.name_scope("Train"):
train_input = PTBInput(config=config, data=train_data, name="TrainInput")
with tf.variable_scope("Model", reuse=None, initializer=initializer):
m = PTBModel(is_training=True, config=config, input_=train_input)
tf.summary.scalar("Training Loss", m.cost)
tf.summary.scalar("Learning Rate", m.lr)
现在跳回ptb_word_lm,在主函数中建立模型,模型的输入函数为PTBInput(),故转去其定义查看
def ptb_producer(raw_data, batch_size, num_steps, name=None):
"""Iterate on the raw PTB data.
This chunks up raw_data into batches of examples and returns Tensors that are drawn from these batches.
Args: raw_data: one of the raw data outputs from ptb_raw_data. batch_size: int, the batch size. num_steps: int, the number of unrolls. name: the name of this operation (optional).
Returns: A pair of Tensors, each shaped [batch_size, num_steps]. The second element of the tuple is the same data time-shifted to the right by one.
Raises: tf.errors.InvalidArgumentError: if batch_size or num_steps are too high. """ with tf.name_scope(name, "PTBProducer", [raw_data, batch_size, num_steps]):
raw_data = tf.convert_to_tensor(raw_data, name="raw_data", dtype=tf.int32)
data_len = tf.size(raw_data)
batch_len = data_len // batch_size
data = tf.reshape(raw_data[0: batch_size * batch_len], [batch_size, batch_len])
epoch_size = (batch_len - 1) // num_steps
assertion = tf.assert_positive(
epoch_size, message="epoch_size == 0, decrease batch_size or num_steps")
with tf.control_dependencies([assertion]):
epoch_size = tf.identity(epoch_size, name="epoch_size")
i = tf.train.range_input_producer(epoch_size, shuffle=False).dequeue()
x = tf.strided_slice(data, [0, i * num_steps], [batch_size, (i + 1) * num_steps])
x.set_shape([batch_size, num_steps])
y = tf.strided_slice(data, [0, i * num_steps + 1], [batch_size, (i + 1) * num_steps + 1])
y.set_shape([batch_size, num_steps])
return x, y
可以很容易看出，其是每次读出一个batch*num_steos的数据。
3.模型部分
当我们看完数据的前处理部分，那么重点就来了，那就是数据的model部分
class PTBModel(object):
"""The PTB model."""
def __init__(self, is_training, config, input_):
self._is_training = is_training
self._input = input_
self._rnn_params = None self._cell = None self.batch_size = input_.batch_size
self.num_steps = input_.num_steps
size = config.hidden_size
vocab_size = config.vocab_size
with tf.device("/cpu:0"):
embedding = tf.get_variable(
"embedding", [vocab_size, size], dtype=data_type())
inputs = tf.nn.embedding_lookup(embedding, input_.input_data)
if is_training and config.keep_prob < 1:
inputs = tf.nn.dropout(inputs, config.keep_prob)
output, state = self._build_rnn_graph(inputs, config, is_training)
softmax_w = tf.get_variable(
"softmax_w", [size, vocab_size], dtype=data_type())
softmax_b = tf.get_variable("softmax_b", [vocab_size], dtype=data_type())
logits = tf.nn.xw_plus_b(output, softmax_w, softmax_b)
# Reshape logits to be a 3-D tensor for sequence loss logits = tf.reshape(logits, [self.batch_size, self.num_steps, vocab_size])
# Use the contrib sequence loss and average over the batches loss = tf.contrib.seq2seq.sequence_loss(
logits, input_.targets, tf.ones([self.batch_size, self.num_steps], dtype=data_type()), average_across_timesteps=False, average_across_batch=True)
# Update the cost self._cost = tf.reduce_sum(loss)
self._final_state = state
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(self._cost, tvars), config.max_grad_norm)
optimizer = tf.train.GradientDescentOptimizer(self._lr)
self._train_op = optimizer.apply_gradients(
zip(grads, tvars), global_step=tf.contrib.framework.get_or_create_global_step())
self._new_lr = tf.placeholder(
tf.float32, shape=[], name="new_learning_rate")
self._lr_update = tf.assign(self._lr, self._new_lr)
def _build_rnn_graph(self, inputs, config, is_training):
if config.rnn_mode == CUDNN:
return self._build_rnn_graph_cudnn(inputs, config, is_training)
else:
return self._build_rnn_graph_lstm(inputs, config, is_training)
def _build_rnn_graph_cudnn(self, inputs, config, is_training):
"""Build the inference graph using CUDNN cell.""" inputs = tf.transpose(inputs, [1, 0, 2])
self._cell = tf.contrib.cudnn_rnn.CudnnLSTM(
num_layers=config.num_layers, num_units=config.hidden_size, input_size=config.hidden_size, dropout=1 - config.keep_prob if is_training else 0)
params_size_t = self._cell.params_size()
self._rnn_params = tf.get_variable(
"lstm_params", initializer=tf.random_uniform(
[params_size_t], -config.init_scale, config.init_scale), validate_shape=False)
c = tf.zeros([config.num_layers, self.batch_size, config.hidden_size], tf.float32)
h = tf.zeros([config.num_layers, self.batch_size, config.hidden_size], tf.float32)
self._initial_state = (tf.contrib.rnn.LSTMStateTuple(h=h, c=c),)
outputs, h, c = self._cell(inputs, h, c, self._rnn_params, is_training)
outputs = tf.transpose(outputs, [1, 0, 2])
outputs = tf.reshape(outputs, [-1, config.hidden_size])
return outputs, (tf.contrib.rnn.LSTMStateTuple(h=h, c=c),)
def _get_lstm_cell(self, config, is_training):
if config.rnn_mode == BASIC:
return tf.contrib.rnn.BasicLSTMCell(
config.hidden_size, forget_bias=0.0, state_is_tuple=True, reuse=not is_training)
if config.rnn_mode == BLOCK:
return tf.contrib.rnn.LSTMBlockCell(
config.hidden_size, forget_bias=0.0)
raise ValueError("rnn_mode %s not supported" % config.rnn_mode)
def _build_rnn_graph_lstm(self, inputs, config, is_training):
"""Build the inference graph using canonical LSTM cells.""" # Slightly better results can be obtained with forget gate biases # initialized to 1 but the hyperparameters of the model would need to be # different than reported in the paper. cell = self._get_lstm_cell(config, is_training)
if is_training and config.keep_prob < 1:
cell = tf.contrib.rnn.DropoutWrapper(
cell, output_keep_prob=config.keep_prob)
cell = tf.contrib.rnn.MultiRNNCell(
[cell for _ in range(config.num_layers)], state_is_tuple=True)
self._initial_state = cell.zero_state(config.batch_size, data_type())
state = self._initial_state
# Simplified version of tensorflow_models/tutorials/rnn/rnn.py's rnn(). # This builds an unrolled LSTM for tutorial purposes only. # In general, use the rnn() or state_saving_rnn() from rnn.py. # # The alternative version of the code below is: # # inputs = tf.unstack(inputs, num=num_steps, axis=1) # outputs, state = tf.contrib.rnn.static_rnn(cell, inputs, # initial_state=self._initial_state) outputs = []
with tf.variable_scope("RNN"):
for time_step in range(self.num_steps):
if time_step > 0: tf.get_variable_scope().reuse_variables()
(cell_output, state) = cell(inputs[:, time_step, :], state)
outputs.append(cell_output)
output = tf.reshape(tf.concat(outputs, 1), [-1, config.hidden_size])
return output, state
def assign_lr(self, session, lr_value):
session.run(self._lr_update, feed_dict={self._new_lr: lr_value})
def export_ops(self, name):
"""Exports ops to collections.""" self._name = name
ops = {util.with_prefix(self._name, "cost"): self._cost}
if self._is_training:
ops.update(lr=self._lr, new_lr=self._new_lr, lr_update=self._lr_update)
if self._rnn_params:
ops.update(rnn_params=self._rnn_params)
for name, op in ops.iteritems():
tf.add_to_collection(name, op)
self._initial_state_name = util.with_prefix(self._name, "initial")
self._final_state_name = util.with_prefix(self._name, "final")
util.export_state_tuples(self._initial_state, self._initial_state_name)
util.export_state_tuples(self._final_state, self._final_state_name)
def import_ops(self):
"""Imports ops from collections.""" if self._is_training:
self._train_op = tf.get_collection_ref("train_op")[0]
self._lr = tf.get_collection_ref("lr")[0]
self._new_lr = tf.get_collection_ref("new_lr")[0]
self._lr_update = tf.get_collection_ref("lr_update")[0]
rnn_params = tf.get_collection_ref("rnn_params")
if self._cell and rnn_params:
params_saveable = tf.contrib.cudnn_rnn.RNNParamsSaveable(
self._cell, self._cell.params_to_canonical, self._cell.canonical_to_params, rnn_params, base_variable_scope="Model/RNN")
tf.add_to_collection(tf.GraphKeys.SAVEABLE_OBJECTS, params_saveable)
self._cost = tf.get_collection_ref(util.with_prefix(self._name, "cost"))[0]
num_replicas = FLAGS.num_gpus if self._name == "Train" else 1 self._initial_state = util.import_state_tuples(
self._initial_state, self._initial_state_name, num_replicas)
self._final_state = util.import_state_tuples(
self._final_state, self._final_state_name, num_replicas)
daixu
