.. _sec_attention-scoring-functions:
注意力评分函数
==============
:numref:`sec_nadaraya-watson`\ 使用了高斯核来对查询和键之间的关系建模。
:eq:`eq_nadaraya-watson-gaussian`\ 中的
高斯核指数部分可以视为\ *注意力评分函数*\ （attention scoring
function）， 简称\ *评分函数*\ （scoring function），
然后把这个函数的输出结果输入到softmax函数中进行运算。
通过上述步骤，将得到与键对应的值的概率分布（即注意力权重）。
最后，注意力汇聚的输出就是基于这些注意力权重的值的加权和。
从宏观来看，上述算法可以用来实现
:numref:`fig_qkv`\ 中的注意力机制框架。
:numref:`fig_attention_output`\ 说明了
如何将注意力汇聚的输出计算成为值的加权和，
其中\ :math:`a`\ 表示注意力评分函数。 由于注意力权重是概率分布，
因此加权和其本质上是加权平均值。
.. _fig_attention_output:
.. figure:: ../img/attention-output.svg
计算注意力汇聚的输出为值的加权和
用数学语言描述，假设有一个查询 :math:`\mathbf{q} \in \mathbb{R}^q`\ 和
:math:`m`\ 个“键－值”对
:math:`(\mathbf{k}_1, \mathbf{v}_1), \ldots, (\mathbf{k}_m, \mathbf{v}_m)`\ ，
其中\ :math:`\mathbf{k}_i \in \mathbb{R}^k`\ ，\ :math:`\mathbf{v}_i \in \mathbb{R}^v`\ 。
注意力汇聚函数\ :math:`f`\ 就被表示成值的加权和：
.. math:: f(\mathbf{q}, (\mathbf{k}_1, \mathbf{v}_1), \ldots, (\mathbf{k}_m, \mathbf{v}_m)) = \sum_{i=1}^m \alpha(\mathbf{q}, \mathbf{k}_i) \mathbf{v}_i \in \mathbb{R}^v,
:label: eq_attn-pooling
其中查询\ :math:`\mathbf{q}`\ 和键\ :math:`\mathbf{k}_i`\ 的注意力权重（标量）
是通过注意力评分函数\ :math:`a`\ 将两个向量映射成标量，
再经过softmax运算得到的：
.. math:: \alpha(\mathbf{q}, \mathbf{k}_i) = \mathrm{softmax}(a(\mathbf{q}, \mathbf{k}_i)) = \frac{\exp(a(\mathbf{q}, \mathbf{k}_i))}{\sum_{j=1}^m \exp(a(\mathbf{q}, \mathbf{k}_j))} \in \mathbb{R}.
:label: eq_attn-scoring-alpha
正如上图所示，选择不同的注意力评分函数\ :math:`a`\ 会导致不同的注意力汇聚操作。
本节将介绍两个流行的评分函数，稍后将用他们来实现更复杂的注意力机制。
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
import math
from mxnet import np, npx
from mxnet.gluon import nn
from d2l import mxnet as d2l
npx.set_np()
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
import math
import torch
from torch import nn
from d2l import torch as d2l
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
import tensorflow as tf
from d2l import tensorflow as d2l
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
import math
import warnings
from d2l import paddle as d2l
warnings.filterwarnings("ignore")
import paddle
from paddle import nn
.. raw:: html
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
def masked_softmax(X, valid_lens):
"""通过在最后一个轴上掩蔽元素来执行softmax操作"""
# X:3D张量，valid_lens:1D或2D张量
if valid_lens is None:
return npx.softmax(X)
else:
shape = X.shape
if valid_lens.ndim == 1:
valid_lens = valid_lens.repeat(shape[1])
else:
valid_lens = valid_lens.reshape(-1)
# 最后一轴上被掩蔽的元素使用一个非常大的负值替换，从而其softmax输出为0
X = npx.sequence_mask(X.reshape(-1, shape[-1]), valid_lens, True,
value=-1e6, axis=1)
return npx.softmax(X).reshape(shape)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
def masked_softmax(X, valid_lens):
"""通过在最后一个轴上掩蔽元素来执行softmax操作"""
# X:3D张量，valid_lens:1D或2D张量
if valid_lens is None:
return nn.functional.softmax(X, dim=-1)
else:
shape = X.shape
if valid_lens.dim() == 1:
valid_lens = torch.repeat_interleave(valid_lens, shape[1])
else:
valid_lens = valid_lens.reshape(-1)
# 最后一轴上被掩蔽的元素使用一个非常大的负值替换，从而其softmax输出为0
X = d2l.sequence_mask(X.reshape(-1, shape[-1]), valid_lens,
value=-1e6)
return nn.functional.softmax(X.reshape(shape), dim=-1)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
def masked_softmax(X, valid_lens):
"""通过在最后一个轴上掩蔽元素来执行softmax操作"""
# X:3D张量，valid_lens:1D或2D张量
if valid_lens is None:
return tf.nn.softmax(X, axis=-1)
else:
shape = X.shape
if len(valid_lens.shape) == 1:
valid_lens = tf.repeat(valid_lens, repeats=shape[1])
else:
valid_lens = tf.reshape(valid_lens, shape=-1)
# 最后一轴上被掩蔽的元素使用一个非常大的负值替换，从而其softmax输出为0
X = d2l.sequence_mask(tf.reshape(X, shape=(-1, shape[-1])),
valid_lens, value=-1e6)
return tf.nn.softmax(tf.reshape(X, shape=shape), axis=-1)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
def masked_softmax(X, valid_lens):
"""通过在最后一个轴上掩蔽元素来执行softmax操作"""
# X:3D张量，valid_lens:1D或2D张量
if valid_lens is None:
return nn.functional.softmax(X, axis=-1)
else:
shape = X.shape
if valid_lens.dim() == 1:
valid_lens = paddle.repeat_interleave(valid_lens, shape[1])
else:
valid_lens = valid_lens.reshape((-1,))
# 最后一轴上被掩蔽的元素使用一个非常大的负值替换，从而其softmax输出为0
X = d2l.sequence_mask(X.reshape((-1, shape[-1])), valid_lens,
value=-1e6)
return nn.functional.softmax(X.reshape(shape), axis=-1)
.. raw:: html
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
masked_softmax(np.random.uniform(size=(2, 2, 4)), np.array([2, 3]))
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
[07:09:32] ../src/storage/storage.cc:196: Using Pooled (Naive) StorageManager for CPU
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
array([[[0.488994 , 0.511006 , 0. , 0. ],
[0.43654838, 0.56345165, 0. , 0. ]],
[[0.28817102, 0.3519408 , 0.3598882 , 0. ],
[0.29034293, 0.25239873, 0.45725834, 0. ]]])
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
masked_softmax(torch.rand(2, 2, 4), torch.tensor([2, 3]))
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
tensor([[[0.5980, 0.4020, 0.0000, 0.0000],
[0.5548, 0.4452, 0.0000, 0.0000]],
[[0.3716, 0.3926, 0.2358, 0.0000],
[0.3455, 0.3337, 0.3208, 0.0000]]])
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
masked_softmax(tf.random.uniform(shape=(2, 2, 4)), tf.constant([2, 3]))
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
masked_softmax(paddle.rand((2, 2, 4)), paddle.to_tensor([2, 3]))
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
Tensor(shape=[2, 2, 4], dtype=float32, place=Place(cpu), stop_gradient=True,
[[[0.34171817, 0.65828186, 0.00000000, 0.00000000],
[0.56181532, 0.43818477, 0.00000000, 0.00000000]],
[[0.27472532, 0.36756957, 0.35770512, 0.00000000],
[0.26933751, 0.40987626, 0.32078618, 0.00000000]]])
.. raw:: html
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
masked_softmax(np.random.uniform(size=(2, 2, 4)),
np.array([[1, 3], [2, 4]]))
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
array([[[1. , 0. , 0. , 0. ],
[0.35848376, 0.36588794, 0.2756283 , 0. ]],
[[0.54370314, 0.45629686, 0. , 0. ],
[0.19598779, 0.25580424, 0.19916737, 0.34904057]]])
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
masked_softmax(torch.rand(2, 2, 4), torch.tensor([[1, 3], [2, 4]]))
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
tensor([[[1.0000, 0.0000, 0.0000, 0.0000],
[0.4125, 0.3273, 0.2602, 0.0000]],
[[0.5254, 0.4746, 0.0000, 0.0000],
[0.3117, 0.2130, 0.1801, 0.2952]]])
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
masked_softmax(tf.random.uniform(shape=(2, 2, 4)), tf.constant([[1, 3], [2, 4]]))
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
masked_softmax(paddle.rand((2, 2, 4)), paddle.to_tensor([[1, 3], [2, 4]]))
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
Tensor(shape=[2, 2, 4], dtype=float32, place=Place(cpu), stop_gradient=True,
[[[1. , 0.00000000, 0.00000000, 0.00000000],
[0.35218778, 0.39164177, 0.25617039, 0.00000000]],
[[0.53993517, 0.46006480, 0.00000000, 0.00000000],
[0.18509592, 0.33311754, 0.25049153, 0.23129500]]])
.. raw:: html
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
class AdditiveAttention(nn.Block):
"""加性注意力"""
def __init__(self, num_hiddens, dropout, **kwargs):
super(AdditiveAttention, self).__init__(**kwargs)
# 使用'flatten=False'只转换最后一个轴，以便其他轴的形状保持不变
self.W_k = nn.Dense(num_hiddens, use_bias=False, flatten=False)
self.W_q = nn.Dense(num_hiddens, use_bias=False, flatten=False)
self.w_v = nn.Dense(1, use_bias=False, flatten=False)
self.dropout = nn.Dropout(dropout)
def forward(self, queries, keys, values, valid_lens):
queries, keys = self.W_q(queries), self.W_k(keys)
# 在维度扩展后，
# queries的形状：(batch_size，查询的个数，1，num_hidden)
# key的形状：(batch_size，1，“键－值”对的个数，num_hiddens)
# 使用广播的方式进行求和
features = np.expand_dims(queries, axis=2) + np.expand_dims(
keys, axis=1)
features = np.tanh(features)
# self.w_v仅有一个输出，因此从形状中移除最后那个维度。
# scores的形状：(batch_size，查询的个数，“键-值”对的个数)
scores = np.squeeze(self.w_v(features), axis=-1)
self.attention_weights = masked_softmax(scores, valid_lens)
# values的形状：(batch_size，“键－值”对的个数，值的维度)
return npx.batch_dot(self.dropout(self.attention_weights), values)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
class AdditiveAttention(nn.Module):
"""加性注意力"""
def __init__(self, key_size, query_size, num_hiddens, dropout, **kwargs):
super(AdditiveAttention, self).__init__(**kwargs)
self.W_k = nn.Linear(key_size, num_hiddens, bias=False)
self.W_q = nn.Linear(query_size, num_hiddens, bias=False)
self.w_v = nn.Linear(num_hiddens, 1, bias=False)
self.dropout = nn.Dropout(dropout)
def forward(self, queries, keys, values, valid_lens):
queries, keys = self.W_q(queries), self.W_k(keys)
# 在维度扩展后，
# queries的形状：(batch_size，查询的个数，1，num_hidden)
# key的形状：(batch_size，1，“键－值”对的个数，num_hiddens)
# 使用广播方式进行求和
features = queries.unsqueeze(2) + keys.unsqueeze(1)
features = torch.tanh(features)
# self.w_v仅有一个输出，因此从形状中移除最后那个维度。
# scores的形状：(batch_size，查询的个数，“键-值”对的个数)
scores = self.w_v(features).squeeze(-1)
self.attention_weights = masked_softmax(scores, valid_lens)
# values的形状：(batch_size，“键－值”对的个数，值的维度)
return torch.bmm(self.dropout(self.attention_weights), values)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
class AdditiveAttention(tf.keras.layers.Layer):
"""Additiveattention."""
def __init__(self, key_size, query_size, num_hiddens, dropout, **kwargs):
super().__init__(**kwargs)
self.W_k = tf.keras.layers.Dense(num_hiddens, use_bias=False)
self.W_q = tf.keras.layers.Dense(num_hiddens, use_bias=False)
self.w_v = tf.keras.layers.Dense(1, use_bias=False)
self.dropout = tf.keras.layers.Dropout(dropout)
def call(self, queries, keys, values, valid_lens, **kwargs):
queries, keys = self.W_q(queries), self.W_k(keys)
# 在维度扩展后，
# queries的形状：(batch_size，查询的个数，1，num_hidden)
# key的形状：(batch_size，1，“键－值”对的个数，num_hiddens)
# 使用广播方式进行求和
features = tf.expand_dims(queries, axis=2) + tf.expand_dims(
keys, axis=1)
features = tf.nn.tanh(features)
# self.w_v仅有一个输出，因此从形状中移除最后那个维度。
# scores的形状：(batch_size，查询的个数，“键-值”对的个数)
scores = tf.squeeze(self.w_v(features), axis=-1)
self.attention_weights = masked_softmax(scores, valid_lens)
# values的形状：(batch_size，“键－值”对的个数，值的维度)
return tf.matmul(self.dropout(
self.attention_weights, **kwargs), values)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
class AdditiveAttention(nn.Layer):
"""加性注意力"""
def __init__(self, key_size, query_size, num_hiddens, dropout, **kwargs):
super(AdditiveAttention, self).__init__(**kwargs)
self.W_k = nn.Linear(key_size, num_hiddens, bias_attr=False)
self.W_q = nn.Linear(query_size, num_hiddens, bias_attr=False)
self.w_v = nn.Linear(num_hiddens, 1, bias_attr=False)
self.dropout = nn.Dropout(dropout)
def forward(self, queries, keys, values, valid_lens):
queries, keys = self.W_q(queries), self.W_k(keys)
# 在维度扩展后，
# queries的形状：(batch_size，查询的个数，1，num_hidden)
# key的形状：(batch_size，1，“键－值”对的个数，num_hiddens)
# 使用广播方式进行求和
features = queries.unsqueeze(2) + keys.unsqueeze(1)
features = paddle.tanh(features)
# self.w_v仅有一个输出，因此从形状中移除最后那个维度。
# scores的形状：(batch_size，查询的个数，“键-值”对的个数)
scores = self.w_v(features).squeeze(-1)
self.attention_weights = masked_softmax(scores, valid_lens)
# values的形状：(batch_size，“键－值”对的个数，值的维度)
return paddle.bmm(self.dropout(self.attention_weights), values)
.. raw:: html
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
queries, keys = np.random.normal(0, 1, (2, 1, 20)), np.ones((2, 10, 2))
# values的小批量数据集中，两个值矩阵是相同的
values = np.arange(40).reshape(1, 10, 4).repeat(2, axis=0)
valid_lens = np.array([2, 6])
attention = AdditiveAttention(num_hiddens=8, dropout=0.1)
attention.initialize()
attention(queries, keys, values, valid_lens)
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
array([[[ 2. , 3. , 4. , 5. ]],
[[10. , 11. , 12.000001, 13. ]]])
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
queries, keys = torch.normal(0, 1, (2, 1, 20)), torch.ones((2, 10, 2))
# values的小批量，两个值矩阵是相同的
values = torch.arange(40, dtype=torch.float32).reshape(1, 10, 4).repeat(
2, 1, 1)
valid_lens = torch.tensor([2, 6])
attention = AdditiveAttention(key_size=2, query_size=20, num_hiddens=8,
dropout=0.1)
attention.eval()
attention(queries, keys, values, valid_lens)
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
tensor([[[ 2.0000, 3.0000, 4.0000, 5.0000]],
[[10.0000, 11.0000, 12.0000, 13.0000]]], grad_fn=)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
queries, keys = tf.random.normal(shape=(2, 1, 20)), tf.ones((2, 10, 2))
# values的小批量，两个值矩阵是相同的
values = tf.repeat(tf.reshape(
tf.range(40, dtype=tf.float32), shape=(1, 10, 4)), repeats=2, axis=0)
valid_lens = tf.constant([2, 6])
attention = AdditiveAttention(key_size=2, query_size=20, num_hiddens=8,
dropout=0.1)
attention(queries, keys, values, valid_lens, training=False)
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
queries, keys = paddle.normal(0, 1, (2, 1, 20)), paddle.ones((2, 10, 2))
# values的小批量，两个值矩阵是相同的
values = paddle.arange(40, dtype=paddle.float32).reshape((1, 10, 4)).tile(
[2, 1, 1])
valid_lens = paddle.to_tensor([2, 6])
attention = AdditiveAttention(key_size=2, query_size=20, num_hiddens=8,
dropout=0.1)
attention.eval()
attention(queries, keys, values, valid_lens)
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
Tensor(shape=[2, 1, 4], dtype=float32, place=Place(cpu), stop_gradient=False,
[[[2. , 3. , 4. , 5. ]],
[[10. , 11. , 12.00000095, 13. ]]])
.. raw:: html
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
d2l.show_heatmaps(attention.attention_weights.reshape((1, 1, 2, 10)),
xlabel='Keys', ylabel='Queries')
.. figure:: output_attention-scoring-functions_2a8fdc_93_0.svg
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
d2l.show_heatmaps(attention.attention_weights.reshape((1, 1, 2, 10)),
xlabel='Keys', ylabel='Queries')
.. figure:: output_attention-scoring-functions_2a8fdc_96_0.svg
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
d2l.show_heatmaps(tf.reshape(attention.attention_weights, (1, 1, 2, 10)),
xlabel='Keys', ylabel='Queries')
.. figure:: output_attention-scoring-functions_2a8fdc_99_0.svg
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
d2l.show_heatmaps(attention.attention_weights.reshape((1, 1, 2, 10)),
xlabel='Keys', ylabel='Queries')
.. figure:: output_attention-scoring-functions_2a8fdc_102_0.svg
.. raw:: html
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
class DotProductAttention(nn.Block):
"""缩放点积注意力"""
def __init__(self, dropout, **kwargs):
super(DotProductAttention, self).__init__(**kwargs)
self.dropout = nn.Dropout(dropout)
# queries的形状：(batch_size，查询的个数，d)
# keys的形状：(batch_size，“键－值”对的个数，d)
# values的形状：(batch_size，“键－值”对的个数，值的维度)
# valid_lens的形状:(batch_size，)或者(batch_size，查询的个数)
def forward(self, queries, keys, values, valid_lens=None):
d = queries.shape[-1]
# 设置transpose_b=True为了交换keys的最后两个维度
scores = npx.batch_dot(queries, keys, transpose_b=True) / math.sqrt(d)
self.attention_weights = masked_softmax(scores, valid_lens)
return npx.batch_dot(self.dropout(self.attention_weights), values)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
class DotProductAttention(nn.Module):
"""缩放点积注意力"""
def __init__(self, dropout, **kwargs):
super(DotProductAttention, self).__init__(**kwargs)
self.dropout = nn.Dropout(dropout)
# queries的形状：(batch_size，查询的个数，d)
# keys的形状：(batch_size，“键－值”对的个数，d)
# values的形状：(batch_size，“键－值”对的个数，值的维度)
# valid_lens的形状:(batch_size，)或者(batch_size，查询的个数)
def forward(self, queries, keys, values, valid_lens=None):
d = queries.shape[-1]
# 设置transpose_b=True为了交换keys的最后两个维度
scores = torch.bmm(queries, keys.transpose(1,2)) / math.sqrt(d)
self.attention_weights = masked_softmax(scores, valid_lens)
return torch.bmm(self.dropout(self.attention_weights), values)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
class DotProductAttention(tf.keras.layers.Layer):
"""Scaleddotproductattention."""
def __init__(self, dropout, **kwargs):
super().__init__(**kwargs)
self.dropout = tf.keras.layers.Dropout(dropout)
# queries的形状：(batch_size，查询的个数，d)
# keys的形状：(batch_size，“键－值”对的个数，d)
# values的形状：(batch_size，“键－值”对的个数，值的维度)
# valid_lens的形状:(batch_size，)或者(batch_size，查询的个数)
def call(self, queries, keys, values, valid_lens, **kwargs):
d = queries.shape[-1]
scores = tf.matmul(queries, keys, transpose_b=True)/tf.math.sqrt(
tf.cast(d, dtype=tf.float32))
self.attention_weights = masked_softmax(scores, valid_lens)
return tf.matmul(self.dropout(self.attention_weights, **kwargs), values)
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
#@save
class DotProductAttention(nn.Layer):
"""缩放点积注意力"""
def __init__(self, dropout, **kwargs):
super(DotProductAttention, self).__init__(**kwargs)
self.dropout = nn.Dropout(dropout)
# queries的形状：(batch_size，查询的个数，d)
# keys的形状：(batch_size，“键－值”对的个数，d)
# values的形状：(batch_size，“键－值”对的个数，值的维度)
# valid_lens的形状:(batch_size，)或者(batch_size，查询的个数)
def forward(self, queries, keys, values, valid_lens=None):
d = queries.shape[-1]
# 设置transpose_b=True为了交换keys的最后两个维度
scores = paddle.bmm(queries, keys.transpose((0,2,1))) / math.sqrt(d)
self.attention_weights = masked_softmax(scores, valid_lens)
return paddle.bmm(self.dropout(self.attention_weights), values)
.. raw:: html
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
queries = np.random.normal(0, 1, (2, 1, 2))
attention = DotProductAttention(dropout=0.5)
attention.initialize()
attention(queries, keys, values, valid_lens)
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
array([[[ 2. , 3. , 4. , 5. ]],
[[10. , 11. , 12.000001, 13. ]]])
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
queries = torch.normal(0, 1, (2, 1, 2))
attention = DotProductAttention(dropout=0.5)
attention.eval()
attention(queries, keys, values, valid_lens)
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
tensor([[[ 2.0000, 3.0000, 4.0000, 5.0000]],
[[10.0000, 11.0000, 12.0000, 13.0000]]])
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
queries = tf.random.normal(shape=(2, 1, 2))
attention = DotProductAttention(dropout=0.5)
attention(queries, keys, values, valid_lens, training=False)
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
queries = paddle.normal(0, 1, (2, 1, 2))
attention = DotProductAttention(dropout=0.5)
attention.eval()
attention(queries, keys, values, valid_lens)
.. raw:: latex
\diilbookstyleoutputcell
.. parsed-literal::
:class: output
Tensor(shape=[2, 1, 4], dtype=float32, place=Place(cpu), stop_gradient=True,
[[[2. , 3. , 4. , 5. ]],
[[10. , 11. , 12.00000095, 13. ]]])
.. raw:: html
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
d2l.show_heatmaps(attention.attention_weights.reshape((1, 1, 2, 10)),
xlabel='Keys', ylabel='Queries')
.. figure:: output_attention-scoring-functions_2a8fdc_138_0.svg
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
d2l.show_heatmaps(attention.attention_weights.reshape((1, 1, 2, 10)),
xlabel='Keys', ylabel='Queries')
.. figure:: output_attention-scoring-functions_2a8fdc_141_0.svg
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
d2l.show_heatmaps(tf.reshape(attention.attention_weights, (1, 1, 2, 10)),
xlabel='Keys', ylabel='Queries')
.. figure:: output_attention-scoring-functions_2a8fdc_144_0.svg
.. raw:: html
.. raw:: html
.. raw:: latex
\diilbookstyleinputcell
.. code:: python
d2l.show_heatmaps(attention.attention_weights.reshape((1, 1, 2, 10)),
xlabel='Keys', ylabel='Queries')
.. figure:: output_attention-scoring-functions_2a8fdc_147_0.svg
.. raw:: html
.. raw:: html
.. raw:: html
`Discussions `__
.. raw:: html
.. raw:: html
`Discussions `__
.. raw:: html
.. raw:: html
`Discussions `__
.. raw:: html
.. raw:: html