调用外部的函数介绍
tf.sampled_softmax_loss()
tf.sampled_softmax_loss()中调用了_compute_sampled_logits() 关于__compute_sampled_logits()
#此函数和nce_loss是差不多的, 取样求loss def sampled_softmax_loss(weights, #[num_classes, dim] biases, #[num_classes] inputs, #[batch_size, dim] labels, #[batch_size, num_true] num_sampled, num_classes, num_true=1, sampled_values=None, remove_accidental_hits=True, partition_strategy="mod", name="sampled_softmax_loss"): #return: [batch_size] 1 2 3 4 5 6 7 8 9 10 11 12 13
关于参数labels:一般情况下,num_true为1, labels的shpae为[batch_size, 1]。假设我们有1000个类别, 使用one_hot形式的label的话, 我们的labels的shape是[batch_size, num_classes]。显然,如果num_classes非常大的话,会影响计算性能。所以,这里采用了一个稀疏的方式,即:使用3代表了[0,0,0,1,0….]
tf.nn.seq2seq.embedding_attention_seq2seq()
创建了input embedding matrix 和 output embedding matrix
def embedding_attention_seq2seq(encoder_inputs, #[T, batch_size] decoder_inputs, #[out_T, batch_size] cell, num_encoder_symbols, num_decoder_symbols, embedding_size, num_heads=1, #只采用一个read head output_projection=None, feed_previous=False, dtype=None, scope=None, initial_state_attention=False): #output_projection: (W, B) W:[output_size, num_decoder_symbols] #B: [num_decoder_symbols] 1 2 3 4 5 6 7 8 9 10 11 12 13 14
(1)这个函数创建了一个inputs 的 embedding matrix.
(2)计算了encoder的 output,并保存起来,用于计算attention
encoder_cell = rnn_cell.EmbeddingWrapper( cell, embedding_classes=num_encoder_symbols, embedding_size=embedding_size)# 创建了inputs的 embedding matrix encoder_outputs, encoder_state = rnn.rnn( encoder_cell, encoder_inputs, dtype=dtype) #return [T ,batch_size,size] 1 2 3 4 5
(3)生成attention states
top_states = [array_ops.reshape(e, [-1, 1, cell.output_size]) for e in encoder_outputs] # T * batch_size * 1 * size attention_states = array_ops.concat(1, top_states) # batch_size*T*size 1 2 3
(4)剩下的工作交给embedding_attention_decoder,embedding_attention_decoder中创建了decoder的embedding matrix
# Decoder. output_size = None if output_projection is None: cell = rnn_cell.OutputProjectionWrapper(cell, num_decoder_symbols) output_size = num_decoder_symbols if isinstance(feed_previous, bool): return embedding_attention_decoder( decoder_inputs, encoder_state, attention_states, cell, num_decoder_symbols, embedding_size, num_heads=num_heads, output_size=output_size, output_projection=output_projection, feed_previous=feed_previous, initial_state_attention=initial_state_attention) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
tf.nn.rnn_cell.EmbeddingWrapper()
embedding_attention_seq2seq中调用了这个类
使用了这个类之后,rnn 的inputs就可以是[batch_size]了,里面保存的是word的id。
此类就是在 cell 前 加了一层embedding
class EmbeddingWrapper(RNNCell): def __init__(self, cell, embedding_classes, embedding_size, initializer=None): def __call__(self, inputs, state, scope=None):#生成embedding矩阵[embedding_classes,embedding_size] #inputs: [batch_size, 1] #return : (output, state) 1 2 3 4 5
tf.nn.rnn_cell.OutputProgectionWrapper()
将rnn_cell的输出映射成想要的维度
class OutputProjectionWrapper(RNNCell): def __init__(self, cell, output_size): # output_size:映射后的size def __call__(self, inputs, state, scope=None): #init 返回一个带output projection的 rnn_cell 1 2 3 4
tf.nn.seq2seq.embedding_attention_decoder()
#生成embedding matrix :[num_symbols, embedding_size]
def embedding_attention_decoder(decoder_inputs, # T*batch_size
initial_state,
attention_states,
cell,
num_symbols,
embedding_size,
num_heads=1,
output_size=None,
output_projection=None,
feed_previous=False,
update_embedding_for_previous=True,
dtype=None,
scope=None,
initial_state_attention=False):
#核心代码
embedding = variable_scope.get_variable("embedding",
[num_symbols, embedding_size]) #output embedding
loop_function = _extract_argmax_and_embed(
embedding, output_projection,
update_embedding_for_previous) if feed_previous else None
emb_inp = [
embedding_ops.embedding_lookup(embedding, i) for i in decoder_inputs]
return attention_decoder(
emb_inp,
initial_state,
attention_states,
cell,
output_size=output_size,
num_heads=num_heads,
loop_function=loop_function,
initial_state_attention=initial_state_attention)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32可以看到,此函数先为 decoder symbols 创建了一个embedding矩阵。然后定义了loop_function。
emb_in是embedded input :[T, batch_size, embedding_size]
函数的主要工作还是交给了attention_decoder()
tf.nn.attention_decoder()
def attention_decoder(decoder_inputs, #[T, batch_size, input_size] initial_state, #[batch_size, cell.states] attention_states, #[batch_size , attn_length , attn_size] cell, output_size=None, num_heads=1, loop_function=None, dtype=None, scope=None, initial_state_attention=False): 1 2 3 4 5 6 7 8 9 10 11
论文中,在计算attention distribution的时候,提到了三个公式
(1)u
t
i
=v
T
∗tanh(W
1
∗h
i
+W
2
∗d
t
)
(1)uit=vT∗tanh(W1∗hi+W2∗dt)
(2)s
t
i
=softmax(a
t
i
)
(2)sit=softmax(ait)
(3)d
′
=∑
i=1
T
A
s
t
i
∗h
i
(3)d′=∑i=1TAsit∗hi
其中,W
1
W1维度是[attn_vec_size, size], h
i
hi:[size,1],我们日常表示输入数据,都是用列向量表示,但是在tensorflow中,趋向用行向量表示。在这个函数中,为了计算 W
1
∗h
i
W1∗hi, 使用了卷积的方式。hidden = array_ops.reshape(
attention_states, [-1, attn_length, 1, attn_size]) #[batch_size * T * 1 * input_size]
hidden_features = []
v = []
attention_vec_size = attn_size # Size of query vectors for attention.
for a in xrange(num_heads):
k = variable_scope.get_variable("AttnW_%d" % a,
[1, 1, attn_size, attention_vec_size])
hidden_features.append(nn_ops.conv2d(hidden, k, [1, 1, 1, 1], "SAME"))
v.append(
variable_scope.get_variable("AttnV_%d" % a, [attention_vec_size])) #attention_vec_size = attn_size
1
2
3
4
5
6
7
8
9
10
11使用conv2d之后,返回的tensor的形状是[batch_size, attn_length, 1, attention_vec_size]
此函数是这么求 W
2
∗d
t
W2∗dt 和 s
i
si的。
y = linear(query, attention_vec_size, True) y = array_ops.reshape(y, [-1, 1, 1, attention_vec_size]) # Attention mask is a softmax of v^T * tanh(...). s = math_ops.reduce_sum( v[a] * math_ops.tanh(hidden_features[a] + y), [2, 3]) #[batch_size, attn_length, 1, attn_size] a = nn_ops.softmax(s) #s" [batch_size * attn_len] # Now calculate the attention-weighted vector d. d = math_ops.reduce_sum( array_ops.reshape(a, [-1, attn_length, 1, 1]) * hidden, [1, 2]) ds.append(array_ops.reshape(d, [-1, attn_size])) 1 2 3 4 5 6 7 8 9 10 11
y=W
2
∗d
t
,d=d
′
y=W2∗dt,d=d′
def rnn()
from tensorflow.python.ops import rnn rnn.rnn() def rnn(cell, inputs, initial_state=None, dtype=None, sequence_length=None, scope=None): #inputs: A length T list of inputs, each a `Tensor` of shape`[batch_size, input_size]` #sequence_length: [batch_size], 指定sample 序列的长度 #return : (outputs, states), outputs: T*batch_size*output_size. states:batch_size*state 1 2 3 4 5 6 7
seq2seqModel
- 创建映射参数 proj_w, proj_b
- 声明:sampled_loss,看了word2vec的就会理解
- 声明:seq2seq_f(),构建了inputs的embedding和outputs的embedding,进行核心计算
- 使用model_with_buckets(),model_with_buckets中调用了seq2seq_f和 sampled_loss
model_with_buckets()
调用方法 tf.nn.seq2seq.model_with_buckets()
def model_with_buckets(encoder_inputs, decoder_inputs, targets, weights, buckets, seq2seq, softmax_loss_function=None, per_example_loss=False, name=None): """Create a sequence-to-sequence model with support for bucketing. The seq2seq argument is a function that defines a sequence-to-sequence model, e.g., seq2seq = lambda x, y: basic_rnn_seq2seq(x, y, rnn_cell.GRUCell(24)) Args: encoder_inputs: A list of Tensors to feed the encoder; first seq2seq input. decoder_inputs: A list of Tensors to feed the decoder; second seq2seq input. targets: A list of 1D batch-sized int32 Tensors (desired output sequence). weights: List of 1D batch-sized float-Tensors to weight the targets. buckets: A list of pairs of (input size, output size) for each bucket. seq2seq: A sequence-to-sequence model function; it takes 2 input that agree with encoder_inputs and decoder_inputs, and returns a pair consisting of outputs and states (as, e.g., basic_rnn_seq2seq). softmax_loss_function: Function (inputs-batch, labels-batch) -> loss-batch to be used instead of the standard softmax (the default if this is None). per_example_loss: Boolean. If set, the returned loss will be a batch-sized tensor of losses for each sequence in the batch. If unset, it will be a scalar with the averaged loss from all examples. name: Optional name for this operation, defaults to "model_with_buckets". Returns: A tuple of the form (outputs, losses), where: outputs: The outputs for each bucket. Its j'th element consists of a list of 2D Tensors. The shape of output tensors can be either [batch_size x output_size] or [batch_size x num_decoder_symbols] depending on the seq2seq model used. losses: List of scalar Tensors, representing losses for each bucket, or, if per_example_loss is set, a list of 1D batch-sized float Tensors. Raises: ValueError: If length of encoder_inputsut, targets, or weights is smaller than the largest (last) bucket. """ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
记住,tensorflow的编码方法是:先构图,再训练。训练是根据feed确定的
