Skip to content

X
X2Paddle

PaddlePaddle / X2Paddle
10 个月 前同步成功

通知 327
Star 698
Fork 167
X
X2Paddle

前往新版Gitcode，体验更适合开发者的 AI 搜索 >>

提交 ede3f485 编写于 作者: J jiangjiajun
浏览文件
操作

update tf2fluid/doc

上级 fe8e5c53
隐藏空白更改
内联 并排
Showing with 101 addition and 205 deletion
tensorflow2fluid/doc/ReadMe.md
浏览文件 @ ede3f485
......@@ -24,9 +24,9 @@
| 18 | [tf.contrib.layers.one_hot_encoding](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/contrib/layers/one_hot_encoding) | [fluid.layers.one_hot](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#one_hot) | 功能一致 |
| 19 | [tf.contrib.layers.softmax](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/contrib/layers/softmax) | [fluid.layers.softmax](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#softmax) | 功能一致 |
| 20 | [tf.contrib.layers.xavier_initializer](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/contrib/layers/xavier_initializer) | [fluid.initializer.Xavier](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/initializer_cn.html#xavier) | 功能一致 |
| 21 | [tf.contrib.rnn.GRUCell](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/contrib/rnn/GRUCell) | [fluid.layers.gru_unit](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#gru_unit) | [差异对比](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.contrib.rnn.GRUCell.md) |
| 22 | [tf.contrib.rnn.MultiRNNCell](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/contrib/rnn/MultiRNNCell) | 无相应接口 | [Paddle实现方法](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.nn.rnn_cell.MultiRNNCell.md) |
| 23 | [tf.contrib.rnn.static_rnn](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/contrib/rnn/static_rnn) | [fluid.layers.DynamicRNN](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#dynamicrnn) | 功能一致 |
| 21 | [tf.nn.rnn.GRUCell](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/rnn_cell/GRUCell) | [fluid.layers.gru_unit](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#gru_unit) | [差异对比](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.contrib.rnn.GRUCell.md) |
| 22 | [tf.nn.rnn.MultiRNNCell](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/rnn_cell/MultiRNNCell) | 无相应接口 | [Paddle实现方法](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.nn.rnn_cell.MultiRNNCell.md) |
| 23 | [tf.nn.rnn.static_rnn](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/static_rnn) | [fluid.layers.DynamicRNN](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#dynamicrnn) | 功能一致 |
| 24 | [tf.convert_to_tensor](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/convert_to_tensor) | [fluid.layers.assign](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#assign) | 功能一致 |
| 25 | [tf.cos](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/cos) | [fluid.layers.cos](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#cos) | 功能一致 |
| 26 | [tf.div](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/div) | [fluid.layers.elementwise_div](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#paddle.fluid.layers.elementwise_div) | 功能一致 |
......@@ -91,7 +91,7 @@
| 85 | [tf.nn.tanh](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/tanh) | [fluid.layers.tanh](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#tanh) | 功能一致 |
| 86 | [tf.one_hot](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/one_hot) | [fluid.layers.one_hot](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#paddle.fluid.layers.one_hot) | [差异对比](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.one_hot.md) |
| 87 | [tf.ones](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/ones) | [fluid.layers.ones](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#ones) | 功能一致 |
| 88 | [tf.ones_initializer](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/ones_initializer) | [fluid.initializer.Constant](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/initializer_cn.html#constant) | 功能一致 |
| 88 | [tf.intializers.ones](https://www.tensorflow.org/versions/r1.14/api_docs/python/tf/initializers/ones) | [fluid.initializer.Constant](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/initializer_cn.html#constant) | 功能一致 |
| 89 | [tf.pad](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/pad) | [fluid.layers.pad](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#pad) | [差异对比](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.pad.md) |
| 90 | [tf.placeholder](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/placeholder) | [fluid.layers.data](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#paddle.fluid.layers.data) | [差异对比](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.placeholder.md) |
| 91 | [tf.pow](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/pow) | [fluid.layers.pow](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#pow) | [差异对比](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.pow.md) |
......@@ -145,4 +145,4 @@
| 139 | [tf.Variable](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/Variable) | [fluid.layers.create_parameter](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#create_parameter) | 功能一致 |
| 140 | [tf.while_loop](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/while_loop) | [fluid.layers.While](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#While) | [差异对比](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.while_loop.md) |
| 141 | [tf.zeros](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/zeros) | [fluid.layers.zeros](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#zeros) | 功能一致 |
| 142 | [tf.zeros_initializer](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/zeros_initializer) | [fluid.initializer.Constant](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/initializer_cn.html#constant) | 功能一致 |
| 142 | [tf.zeros_initializer](https://www.tensorflow.org/versions/r1.14/api_docs/python/tf/zeros_initializer) | [fluid.initializer.Constant](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/initializer_cn.html#constant) | 功能一致 |
tensorflow2fluid/doc/compare_op.md
浏览文件 @ ede3f485
......@@ -4,8 +4,8 @@
| TensorFlow接口 | PaddlePaddle接口 |
|--------------------------|-------------------------------------------------|
|[tf.math.less_equal](https://www.tensorflow.org/api_docs/python/tf/math/less_equal)|运算符`<=`|
|[tf.math.greater](https://www.tensorflow.org/api_docs/python/tf/math/greater)|运算符`>`|
|[tf.math.greater_equal](https://www.tensorflow.org/api_docs/python/tf/math/greater_equal)|运算符`>=`|
|[tf.math.equal](https://www.tensorflow.org/api_docs/python/tf/math/equal)|运算符`==` 或 [paddle.fluid.layers.equal](http://paddlepaddle.org/documentation/docs/zh/1.3/api_cn/layers_cn.html#permalink-7-equal) |
|[tf.math.less](https://www.tensorflow.org/api_docs/python/tf/math/less)|运算符`<` 或 [paddle.fluid.layers.less_than](http://paddlepaddle.org/documentation/docs/zh/1.3/api_cn/layers_cn.html#permalink-11-less_than) |
|[tf.math.less_equal](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/less_equal)|运算符`<=`|
|[tf.math.greater](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/greater)|运算符`>`|
|[tf.math.greater_equal](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/greater_equal)|运算符`>=`|
|[tf.math.equal](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/equal)|运算符`==` 或 [paddle.fluid.layers.equal](http://paddlepaddle.org/documentation/docs/zh/1.3/api_cn/layers_cn.html#permalink-7-equal) |
|[tf.math.less](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/less)|运算符`<` 或 [paddle.fluid.layers.less_than](http://paddlepaddle.org/documentation/docs/zh/1.3/api_cn/layers_cn.html#permalink-11-less_than) |
\ No newline at end of file
tensorflow2fluid/doc/tf.case.md
浏览文件 @ ede3f485
## tf.case
### [tf.case](https://www.tensorflow.org/api_docs/python/tf/case)
### [tf.case](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/case)
```python
tf.case(
......@@ -52,4 +52,4 @@ with fluid.layers.control_flow.Switch() as switch:
with switch.default():
fluid.layers.tensor.assign(input=lr_0, output=lr)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.clip_by_global_norm.md
浏览文件 @ ede3f485
## tf.clip_by_global_norm
### [tf.clip_by_global_norm](https://www.tensorflow.org/api_docs/python/tf/clip_by_global_norm)
### [tf.clip_by_global_norm](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/clip_by_global_norm)
```python
tf.clip_by_global_norm(
......@@ -46,4 +46,4 @@ with fluid.program_guard(main_program=prog_clip):
# 执行裁剪并获取结果
p_g_clip = fluid.clip.append_gradient_clip_ops(p_g_clip)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.clip_by_norm.md
浏览文件 @ ede3f485
## tf.clip_by_norm
### [tf.clip_by_norm](https://www.tensorflow.org/api_docs/python/tf/clip_by_norm)
### [tf.clip_by_norm](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/clip_by_norm)
``` python
tf.clip_by_norm(
......@@ -24,4 +24,4 @@ paddle.fluid.layers.clip_by_norm(
#### 计算方式
TensorFlow: 使用参数`axis`指定的轴计算L2范数`l2-norm`，如若`axis`为None，则表示使用整个输入数据的L2范数；
PaddlePaddle：使用整个输入数据的L2范数。
PaddlePaddle：使用整个输入数据的L2范数。
\ No newline at end of file
tensorflow2fluid/doc/tf.contrib.layers.flatten.md
浏览文件 @ ede3f485
## tf.contrib.layers.flatten
### [tf.contrib.layers.flatten](https://www.tensorflow.org/api_docs/python/tf/contrib/layers/flatten)
### [tf.contrib.layers.flatten](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/contrib/layers/flatten)
```python
tf.contrib.layers.flatten(
......@@ -34,4 +34,4 @@ PaddlePaddle：使用`axis`指定两次合并的维度边界，参考下面示
out = fluid.layers.flatten(x, axis=2)
out.shape # [2*3, 4*5]
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.contrib.rnn.GRUCell.md 已删除 100644 → 0
浏览文件 @ fe8e5c53
## tf.contrib.rnn.GRUCell
### [tf.contrib.rnn.GRUCell](https://www.tensorflow.org/api_docs/python/tf/nn/rnn_cell/GRUCell)
```python
tf.contrib.rnn.GRUCell(
num_units,
activation=None,
reuse=None,
kernel_initializer=None,
bias_initializer=None,
name=None,
dtype=None,
**kwargs
)
```
### [paddle.fluid.layers.gru_unit](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#gru-unit)
```python
paddle.fluid.layers.gru_unit(
input,
hidden,
size,
param_attr=None,
bias_attr=None,
activation='tanh',
gate_activation='sigmoid',
origin_mode=False
)
```
### 功能差异
#### 实现方式
TensorFlow：GRU的实现方式见论文[Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation](http://arxiv.org/abs/1406.1078)；
PaddlePaddle：GRU有两种实现方式，当设置`origin_mode=False`时，与TensorFlow实现方式一致；当设置`origin_mode=True`时，实现方式则参考论文[Empirical Evaluation of
Gated Recurrent Neural Networks
on Sequence Modeling](https://arxiv.org/pdf/1412.3555.pdf)。
#### 使用方式
TensorFlow：首先定义`GRUCell`对象，定义对象时只需要指定单元数`num_units`；由于`GRUCell`内部定义了`__call__`方法，因而其对象是可调用对象，直接使用`step_output, cur_state = cell(step_input, last_state)`的形式，可以计算得到当前步的输出与状态;
PaddlePaddle：提供op形式的调用接口，通常与[paddle.fluid.layers.DynamicRNN](http://paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#dynamicrnn)配合使用，以获取序列中的单步输入。**注意，为了提高`gru_unit`的计算效率，用户在使用该接口时需要遵从如下约定：假设要指定的GRU单元数为`num_units`，则`size`以及`input.shape[-1]`必须为`3*num_units`，`hidden.shape[-1]`为`num_units`，见如下代码示例小节。**
#### 返回值
TensorFlow：返回一个二元组，分别是当前时刻的输出值与隐藏状态，实际上输出值与隐藏状态为相同的tensor；
PaddlePaddle：返回一个三元组，即`(hidden_value, reset_hidden_value, gate_value)`。后面两个元素为内部使用，用户可以只关注第一个元素。
### 代码示例
```
emb_size = 32
emb_vocab = 10000
num_unit_0 = 10
data = fluid.layers.data(name='input', shape=[1], dtype='int64', lod_level=1)
embedding = fluid.layers.embedding(input=data, size=[emb_vocab, emb_size],
is_sparse=False)
# 为了调用gru_unit，输入最后的维度必须为实际单元数的3倍
emb_fc = layers.fc(embedding, num_unit_0 * 3)
drnn = fluid.layers.DynamicRNN()
with drnn.block():
word = drnn.step_input(emb_fc)
# 指定上一时刻的隐状态，单元数为num_unit_0
prev_hid0 = drnn.memory(shape=[num_unit_0])
# 执行gru_unit计算，num_unit_0 为实际的单元数
cur_hid0, _, _ = layers.gru_unit(word, prev_hid0, num_unit_0 * 3)
# 更新隐状态
drnn.update_memory(prev_hid0, cur_hid0)
drnn.output(cur_hid0)
out = drnn()
last = fluid.layers.sequence_last_step(out)
```
tensorflow2fluid/doc/tf.expand_dims.md
浏览文件 @ ede3f485
## tf.expand_dims
### [tf.expand_dims](https://www.tensorflow.org/api_docs/python/tf/expand_dims)
### [tf.expand_dims](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/expand_dims)
``` python
tf.expand_dims(
input,
......@@ -39,5 +38,4 @@ out = fluid.layers.unsqueeze(t, [-1])
# 输出 tensor out 的 shape 为[1, 1，3, 4]
out = fluid.layers.unsqueeze(t, [0, 1])
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.image.non_max_suppression.md
浏览文件 @ ede3f485
## tf.image.non_max_suppression
### [tf.image.non_max_suppression](https://www.tensorflow.org/api_docs/python/tf/image/non_max_suppression)
### [tf.image.non_max_suppression](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/image/non_max_suppression)
``` python
tf.image.non_max_suppression(
boxes,
......@@ -54,4 +53,4 @@ selected_boxes = fluid.layers.multiclass_nms(
nms_top_k=-1,
keep_top_k=300,
nms_threshold=0.7)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.image.resize_images.md
浏览文件 @ ede3f485
## tf.image.resize_images
### [tf.image.resize_images](https://www.tensorflow.org/api_docs/python/tf/image/resize_images)
### [tf.image.resize_images](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/image/resize_images)
``` python
tf.image.resize_images(
images,
......@@ -38,5 +37,4 @@ inputs = fluid.layers.data(dtype='float32', shape=[3, 300, 300], name='inputs')
# 输出shape为[3, 400, 500]
outputs = fluid.layers.image_reisze(inputs, [400, 500])
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.layers.conv2d.md
浏览文件 @ ede3f485
## tf.layers.conv2d
### [tf.layers.conv2d](https://www.tensorflow.org/api_docs/python/tf/layers/conv2d)
### [tf.layers.conv2d](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/layers/conv2d)
``` python
tf.layers.conv2d(
inputs,
......@@ -73,7 +72,7 @@ pad_right = pad_size - pad_left
PaddlePaddle：`padding`参数表示在输入图像四周padding的size大小。
#### 参数差异
TensorFlow：深度可分离卷积使用[tf.layers.separable_conv2d](https://www.tensorflow.org/api_docs/python/tf/layers/separable_conv2d)接口;
TensorFlow：深度可分离卷积使用[tf.layers.separable_conv2d](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/layers/separable_conv2d)接口;
PaddlePaddle: 使用`paddle.fluid.layers.conv2d`，可参考
[PaddlePaddle对卷积的说明文档](http://paddlepaddle.org/documentation/docs/zh/1.4/api_guides/low_level/layers/conv.html), 同时也可参考[tf.nn.separable_conv2d](https://github.com/PaddlePaddle/X2Paddle/blob/master/tensorflow2fluid/doc/tf.nn.separable_conv2d.md)中的代码示例。
......@@ -85,4 +84,4 @@ PaddlePaddle: 使用`paddle.fluid.layers.conv2d`，可参考
# 卷积核Shape: (5, 3, 4, 4)
inputs = paddle.fluid.layers.data(dtype='float32', shape=[3, 200, 200], name='inputs)
pad_inputs = paddle.fluid.layers.pad2d(inputs, paddings=[1, 2, 1, 2])
outputs = paddle.fluid.layers.conv2d(pad_inputs, 5, [4, 4], (1, 1))
outputs = paddle.fluid.layers.conv2d(pad_inputs, 5, [4, 4], (1, 1))
\ No newline at end of file
tensorflow2fluid/doc/tf.layers.dense.md
浏览文件 @ ede3f485
## tf.layers.dense
### [tf.layers.dense](https://www.tensorflow.org/api_docs/python/tf/layers/dense)
### [tf.layers.dense](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/layers/dense)
``` python
tf.layers.dense(
inputs,
......@@ -63,4 +62,4 @@ out = fluid.layers.fc(t, size=6, \
# size=6, num_flatten_dims=2，输出tensor的shape为[2, 3, 6]
out = fluid.layers.fc(t, size=6, num_flatten_dims=2)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.losses.mean_and_squared_error.md
浏览文件 @ ede3f485
## tf.losses.mean_and_squared_error
### [tf.losses.mean_and_squared_error](https://www.tensorflow.org/api_docs/python/tf/losses/mean_squared_error)
### [tf.losses.mean_and_squared_error](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/losses/mean_squared_error)
``` python
tf.losses.mean_squared_error(
......@@ -25,4 +25,4 @@ paddle.fluid.layers.square_error_cost(
#### 计算方式
TensorFlow: 提供`weights`参数，通过传入`weights`参数的shape，可实现不同的加权方式；
PaddlePaddle：不支持加权。
PaddlePaddle：不支持加权。
\ No newline at end of file
tensorflow2fluid/doc/tf.losses.sigmoid_cross_entropy.md
浏览文件 @ ede3f485
## tf.losses.sigmoid_cross_entropy
### [tf.losses.sigmoid_cross_entropy](https://www.tensorflow.org/api_docs/python/tf/losses/sigmoid_cross_entropy)
### [tf.losses.sigmoid_cross_entropy](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/losses/sigmoid_cross_entropy)
```python
tf.losses.sigmoid_cross_entropy(
......@@ -53,4 +53,4 @@ PaddlePaddle：通过设置`normalize`，各样本损失函数会除以除去`ig
out = fluid.layers.sigmoid_cross_entropy_with_logits(x, label)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.math.is_finite.md
浏览文件 @ ede3f485
## tf.math.is_finite
### [tf.math.is_finite](https://www.tensorflow.org/api_docs/python/tf/math/is_finite)
### [tf.math.is_finite](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/is_finite)
``` python
tf.math.is_finite(
x,
......@@ -31,4 +30,4 @@ result = tf.is_finite(inputs)
# 输入[2.1, 3.2, 4.5]
# 输出True
result = fluid.layers.isfinite(inputs)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.math.rsqrt.md
浏览文件 @ ede3f485
## tf.math.rsqrt
### [tf.math.rsqrt](https://www.tensorflow.org/api_docs/python/tf/math/rsqrt)
### [tf.math.rsqrt](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/rsqrt)
``` python
tf.math.rsqrt(
x,
......@@ -23,4 +23,4 @@ inputs = fluid.layers.data(dtype='float32', shape=[1000], name='inputs')
# 调用上述自定义函数
result = rsqrt(inputs)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.matmul.md
浏览文件 @ ede3f485
## tf.matmul
### [tf.matmul](https://www.tensorflow.org/api_docs/python/tf/linalg/matmul)
### [tf.matmul](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/linalg/matmul)
``` python
tf.matmul(
a,
......@@ -60,4 +59,4 @@ fluid.layers.matmul(x, y) # out: [1]
# x: [M], y: [N]
fluid.layers.matmul(x, y, True, True) # out: [M, N]
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.avg_pool.md
浏览文件 @ ede3f485
## tf.nn.avg_pool
### [tf.nn.avg_pool](https://www.tensorflow.org/versions/r1.10/api_docs/python/tf/nn/avg_pool)
### [tf.nn.avg_pool](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/avg_pool)
``` python
tf.nn.avg_pool(
......
tensorflow2fluid/doc/tf.nn.bidirectional_dynamic_rnn.md
浏览文件 @ ede3f485
## tf.nn.bidirectional_dynamic_rnn
### [tf.nn.bidirectional_dynamic_rnn](https://www.tensorflow.org/api_docs/python/tf/nn/bidirectional_dynamic_rnn)
### [tf.nn.bidirectional_dynamic_rnn](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/bidirectional_dynamic_rnn)
```python
tf.nn.bidirectional_dynamic_rnn(
......@@ -71,4 +71,4 @@ rev_rev_out = fluid.layers.sequence_reverse(rev_out)
# 合并得到最后的输出，其shape为(-1, 32)
concat_out = layers.concat([out, rev_rev_out], axis=1)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.conv2d.md
浏览文件 @ ede3f485
## tf.nn.conv2d
### [tf.nn.conv2d](https://www.tensorflow.org/api_docs/python/tf/nn/conv2d)
### [tf.nn.conv2d](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/conv2d)
```python
tf.nn.conv2d(
......@@ -48,4 +48,4 @@ create_kernel = fluid.layers.create_parameters(shape=[5, 3, 2, 2], dtype='float3
# PaddlePaddle中可通过相同的参数命名引用同一个参数变量
# 通过指定卷积核参数名(param_attr)为'kernel'，引用了create_kernel
result = fluid.layers.conv2d(inputs, 5, [2, 2], param_attr='kernel')
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.conv2d_transpose.md
浏览文件 @ ede3f485
## tf.nn.conv2d_transpose
### [tf.nn.conv2d_transpose](https://www.tensorflow.org/api_docs/python/tf/nn/conv2d_transpose)
### [tf.nn.conv2d_transpose](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/conv2d_transpose)
``` python
tf.nn.conv2d_transpose(
value,
......@@ -93,4 +92,4 @@ inputs = fluid.layers.data(dtype='float32', shape=[3, 20, 20], name='inputs)
outputs = fluid.layers.conv2d_transpose(pad_inputs, 3, filter_size=[5, 5],
padding=[1, 1], stride=[2, 2], bias_attr=False)
# 裁剪后结果即为与TensorFlow一致
outputs = fluid.layers.crop(outputs, shape=[-1, 3, 40, 40])
outputs = fluid.layers.crop(outputs, shape=[-1, 3, 40, 40])
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.conv3d_transpose.md
浏览文件 @ ede3f485
## tf.nn.conv3d_transpose
### [tf.nn.conv3d_transpose](https://www.tensorflow.org/api_docs/python/tf/nn/conv3d_transpose)
### [tf.nn.conv3d_transpose](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/conv3d_transpose)
``` python
tf.nn.conv3d_transpose(
value,
......@@ -93,4 +92,4 @@ inputs = fluid.layers.data(dtype='float32', shape=[3, 5, 20, 40], name='inputs)
outputs = fluid.layers.conv3d(inputs, 7, filter_size=(2, 4, 5), stride=(1, 2, 2),
padding=(0, 1, 1), bias_attr=False)
# 裁剪后结果即为与TensorFlow一致
outputs = fluid.layers.crop(outputs, shape=[-1, 7, 5, 40, 80])
outputs = fluid.layers.crop(outputs, shape=[-1, 7, 5, 40, 80])
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.depthwise_conv2d.md
浏览文件 @ ede3f485
## tf.nn.depthwise_conv2d
### [tf.nn.depthwise_conv2d](https://www.tensorflow.org/api_docs/python/tf/nn/depthwise_conv2d)
### [tf.nn.depthwise_conv2d](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/depthwise_conv2d)
```python
tf.nn.depthwise_conv2d(
......@@ -84,4 +84,4 @@ inputs = fluid.layers.data(dtype='float32', shape=[3, 20, 20], name='inputs')
inputs = fluid.layers.pad2d(inputs, paddings=[1, 2, 1, 2])
#输出shape:[-1, 3, 20, 20]
result = fluid.layers.conv2d(inputs, 3, filter_size=[4, 4], groups=3, bias_attr=False)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.dropout.md
浏览文件 @ ede3f485
## tf.dropout
### [tf.nn.dropout](https://www.tensorflow.org/api_docs/python/tf/nn/dropout)
### [tf.nn.dropout](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/dropout)
``` python
tf.nn.dropout(
x,
......@@ -47,4 +46,4 @@ out = fluid.layers.dropout(t, dropout_prob=0.2, dropout_implementation="upscale_
# inference 时关闭dropout
inference_program = fluid.default_main_program().clone(for_test=True)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.dynamic_rnn.md
浏览文件 @ ede3f485
## tf.nn.dynamic_rnn
### [tf.nn.dynamic_rnn](https://www.tensorflow.org/api_docs/python/tf/nn/dynamic_rnn)
### [tf.nn.dynamic_rnn](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/dynamic_rnn)
``` python
tf.nn.dynamic_rnn(
cell,
......@@ -75,4 +75,4 @@ state = fluid.layers.sequence_last_step(outputs)
为了简化用户定义动态RNN的过程，paddle有如下op可供选择：
- [paddle.fluid.layers.dynamic_lstm](http://www.paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#dynamic-lstm)：相当于 `tf.nn.dynamic_rnn`结合`tf.nn.rnn_cell.LSTMCell`
- [paddle.fluid.layers.dynamic_gru](http://www.paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#dynamic-gru)：相当于`tf.nn.dynamic_rnn`结合`tf.nn.rnn_cell.GRUCell`
- [paddle.fluid.layers.dynamic_gru](http://www.paddlepaddle.org/documentation/docs/zh/1.4/api_cn/layers_cn.html#dynamic-gru)：相当于`tf.nn.dynamic_rnn`结合`tf.nn.rnn_cell.GRUCell`
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.l2_normalize.md
浏览文件 @ ede3f485
## tf.nn.l2_normalize
### [tf.nn.l2_normalize](https://www.tensorflow.org/api_docs/python/tf/math/l2_normalize)
### [tf.nn.l2_normalize](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/l2_normalize)
```python
tf.math.l2_normalize(
......@@ -38,4 +37,4 @@ PaddlePaddle：计算方式为`output = x / sqrt(sum(x^2) + epsilon))`。
# out同样是shape[3,2]的张量，axis设置为1，表示将x中每个行向量做归一化
out = fluid.layers.l2_normalize(x, axis=1)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.lrn.md
浏览文件 @ ede3f485
## tf.nn.lrn
### [tf.nn.lrn](https://www.tensorflow.org/api_docs/python/tf/nn/local_response_normalization)
### [tf.nn.lrn](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/local_response_normalization)
```python
tf.nn.local_response_normalization(
......@@ -38,4 +37,4 @@ $$output(i,x,y)=input(i,x,y)/(k+\alpha\sum_{j=max(0,i-\frac{n}{2})}^{min(C,i+\fr
#### 输入格式
TensorFlow: 默认输入`NHWC`格式数据；
PaddlePaddle: 默认输入`NCHW`格式数据，
PaddlePaddle: 默认输入`NCHW`格式数据，
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.max_pool.md
浏览文件 @ ede3f485
## tf.nn.max_pool
### [tf.nn.max_pool](https://www.tensorflow.org/api_docs/python/tf/nn/max_pool)
### [tf.nn.max_pool](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/max_pool)
``` python
tf.nn.max_pool(
......@@ -51,4 +51,4 @@ inputs = fluid.layers.data(dtype='float32', shape=[3, 300, 300], name='inputs')
# 在最右、最下进行padding
pad_res = fluid.layers.pad2d(inputs, padding=[0, 1, 0, 1])
conv_res = fluid.layers.pool2d(pad_res, pool_size=3, pool_type='max', pool_stride=2)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.reduce_logsumexp.md
浏览文件 @ ede3f485
## tf.math.reduce_logsumexp
### [tf.math.reduce_logsumexp](https://www.tensorflow.org/api_docs/python/tf/math/reduce_logsumexp)
### [tf.math.reduce_logsumexp](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/reduce_logsumexp)
``` python
tf.math.log_softmax(
logits,
......@@ -26,4 +26,4 @@ inputs = fluid.layers.data(dtype='float32', shape=[1000], name='inputs')
# 调用上述自定义函数
result = reduce_logsumexp(inputs)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.rnn_cell.LSTMCell.md
浏览文件 @ ede3f485
## tf.nn.rnn_cell.LSTMCell
### [tf.nn.rnn_cell.LSTMCell](https://www.tensorflow.org/api_docs/python/tf/nn/rnn_cell/LSTMCell)
### [tf.nn.rnn_cell.LSTMCell](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/rnn_cell/LSTMCell)
```python
tf.nn.rnn_cell.LSTMCell(
......@@ -85,4 +85,4 @@ out = drnn()
# 获取最后时刻的输出
last = fluid.layers.sequence_last(out)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.rnn_cell.MultiRNNCell.md
浏览文件 @ ede3f485
## tf.nn.rnn_cell.MultiRNNCell
### [tf.nn.rnn_cell.MultiRNNCell](https://www.tensorflow.org/api_docs/python/tf/nn/rnn_cell/MultiRNNCell)
### [tf.nn.rnn_cell.MultiRNNCell](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/rnn_cell/MultiRNNCell)
```python
tf.nn.rnn_cell.MultiRNNCell(
__init__(
cells,
state_is_tuple=True
)
......
tensorflow2fluid/doc/tf.nn.separable_conv2d.md
浏览文件 @ ede3f485
## tf.nn.separable_conv2d
### [tf.nn.separable_conv2d](https://www.tensorflow.org/api_docs/python/tf/nn/separable_conv2d)
### [tf.nn.separable_conv2d](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/separable_conv2d)
``` python
tf.nn.separable_conv2d(
input,
......@@ -30,4 +30,4 @@ depthwise_result = fluid.layers.conv2d(input, 3, filter_size=[4, 4],
pointwise_result = fluid.layers.conv2d(depthwise_result, filter_size=[1, 1],
stride=[1, 1], bias_attr=False)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.nn.softmax_cross_entropy_with_logits.md
浏览文件 @ ede3f485
## tf.nn.softmax_cross_entropy_with_logits
### [tf.nn.rnn_cell.MultiRNNCell](https://www.tensorflow.org/api_docs/python/tf/nn/softmax_cross_entropy_with_logits)
### [tf.nn.rnn_cell.MultiRNNCell](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/softmax_cross_entropy_with_logits)
```python
tf.nn.softmax_cross_entropy_with_logits(
......@@ -20,7 +20,8 @@ paddle.fluid.layers.softmax_with_cross_entropy(
soft_label=False,
ignore_index=-100,
numeric_stable_mode=False,
return_softmax=False
return_softmax=False,
axis=-1
)
```
......
tensorflow2fluid/doc/tf.nn.top_k.md
浏览文件 @ ede3f485
## tf.nn.top_k
### [tf.nn.top_k](https://www.tensorflow.org/api_docs/python/tf/nn/top_k)
### [tf.nn.top_k](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/nn/top_k)
``` python
tf.math.top_k(
input,
......@@ -32,4 +31,4 @@ PaddlePaddle: 对返回的top-k tensor进行降序排序；`k`没有默认值，
# 当k=2时，输出 tensor out 为[[6,3], [8,3]]，index为[[1,2],[2,0]]
out, index = fluid.layers.topk(t, k=1)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.one_hot.md
浏览文件 @ ede3f485
## tf.one_hot
### [tf.one_hot](https://www.tensorflow.org/api_docs/python/tf/one_hot)
### [tf.one_hot](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/one_hot)
``` python
tf.one_hot(
indices,
......@@ -38,4 +37,4 @@ PaddlePaddle：无对应配置选项，即为默认的`1`和`0`。
# depth 为3时，输出 tensor out 为[[0, 1, 0], [0, 0, 1]]
out = fluid.layers.one_hot(t, 3)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.pad.md
浏览文件 @ ede3f485
## tf.pad
### [tf.pad](https://www.tensorflow.org/api_docs/python/tf/pad)
### [tf.pad](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/pad)
``` python
tf.pad(
tensor,
......@@ -34,4 +33,4 @@ PaddlePaddle：目前仅支持采用常量进行padding；指定padding长度时
# 第0维前面padding长度为0，后面padding长度为1；第1维前面padding长度为1，后面padding长度为2
out = fluid.layers.pad(t, paddings=[0,1,1,2])
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.placeholder.md
浏览文件 @ ede3f485
## tf.placeholder
### [tf.placeholder](https://www.tensorflow.org/api_docs/python/tf/placeholder)
### [tf.placeholder](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/placeholder)
``` python
tf.placeholder(
dtype,
......@@ -36,4 +35,4 @@ out = fluid.layers.data('out', shape=[3, 4], dtype='float32')
# 创建输入型tensor out，其shape为[3, -1, 4], 数据类型为float32
out = fluid.layers.data('out', shape=[3, -1, 4], append_batch_size=False, dtype='float32')
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.pow.md
浏览文件 @ ede3f485
## tf.pow
### [tf.pow](https://www.tensorflow.org/api_docs/python/tf/math/pow)
### [tf.pow](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/pow)
```python
tf.math.pow(
......@@ -33,4 +33,4 @@ PaddlePaddle：`x`为tensor，`factor`为浮点数，返回值为`x`每个元素
# x为张量 [2, 3]
out = fluid.layers.pow(x, 2.0) # [4，9]
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.print.md
浏览文件 @ ede3f485
## tf.print
### [tf.print](https://www.tensorflow.org/api_docs/python/tf/print)
### [tf.print](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/print)
```python
tf.print(
......@@ -47,4 +47,4 @@ PaddlePaddle：通过设置`print_phase`，可以控制是否打印`input`的梯
# 打印input的内容，如果有梯度的话也将打印梯度
print(input, message="content of input")
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.reshape.md
浏览文件 @ ede3f485
## tf.reshape
### [tf.reshape](https://www.tensorflow.org/api_docs/python/tf/reshape)
### [tf.reshape](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/reshape)
``` python
tf.reshape(
tensor,
......@@ -37,5 +36,4 @@ out = fluid.layers.reshape(t, [-1, 6])
# 输出 tensor out 的 shape 为[3, 2, 2]
out = fluid.layers.reshape(t, [0, 2, 2])
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.reverse_sequence.md
浏览文件 @ ede3f485
## tf.reverse_sequence
### [tf.reverse_sequence](https://www.tensorflow.org/api_docs/python/tf/reverse_sequence)
### [tf.reverse_sequence](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/reverse_sequence)
```python
tf.reverse_sequence(
......@@ -44,4 +44,4 @@ PaddlePaddle：由于`LoDTensor`本身已经携带序列信息，因而不需要
# out[0:2, 6] = x[2:0:-1, 6]
# out[2:5, 6] = x[5:2:-1, 6]
out = fluid.layers.sequence_reverse(x)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.scatter_update.md
浏览文件 @ ede3f485
## tf.scatter_update
### [tf.scatter_update](https://www.tensorflow.org/api_docs/python/tf/scatter_update)
### [tf.scatter_update](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/scatter_update)
```python
tf.scatter_update(
......@@ -19,7 +19,8 @@ paddle.fluid.layers.scatter(
input,
index,
updates,
name=None
name=None,
overwrite=True
)
```
......@@ -32,7 +33,7 @@ PaddlePaddle：`index`只支持1-d Variable。
#### 其他
Tensorflow：`updates`支持numpy-style broadcasting;
PaddlePaddle：`updates`要求其rank与`input`相同，同时`updates.shape[0]`等于`index.shape[0]`。
PaddlePaddle：`updates`要求其rank与`input`相同，同时`updates.shape[0]`等于`index.shape[0]`。此外`overwrite`参数提供了当存在重复index时，两种不同的梯度更新策略。
### 代码示例
```
......
tensorflow2fluid/doc/tf.slice.md
浏览文件 @ ede3f485
## tf.slice
### [tf.slice](https://www.tensorflow.org/api_docs/python/tf/slice)
### [tf.slice](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/slice)
``` python
tf.slice(
input_,
......@@ -40,4 +39,4 @@ out = fluid.layers.slice(t, axes=[0,1], starts=[0,1], ends=[2,3])
# 输出 tensor out 为[[1,2],[5,6],[9,10]]
out = fluid.layers.slice(t, axes=[1], starts=[1], ends=[3])
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.split.md
浏览文件 @ ede3f485
## tf.split
### [tf.split](https://www.tensorflow.org/api_docs/python/tf/split)
### [tf.split](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/split)
```python
tf.split(
......@@ -39,4 +39,4 @@ x0.shape # [3, 3, 5]
x1.shape # [3, 3, 5]
x2.shape # [3, 3, 5]
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.squared_difference.md
浏览文件 @ ede3f485
## tf.squared_difference
### [tf.squared_diffenrece](https://www.tensorflow.org/api_docs/python/tf/math/squared_difference)
### [tf.squared_diffenrece](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/math/squared_difference)
``` python
tf.math.squared_difference(
x,
......@@ -24,4 +24,4 @@ input_x = fluid.layers.data(dtype='float32', shape=[1000], name='input_x')
input_y = fluid.layers.data(dtype='float32', shape=[1000], name='input_y')
# 调用上述自定义函数
result = squared_difference(input_x, input_y)
```
```
\ No newline at end of file
tensorflow2fluid/doc/tf.stop_gradient.md
浏览文件 @ ede3f485
## tf.stop_gradient
### [tf.stop_gradient](https://www.tensorflow.org/api_docs/python/tf/stop_gradient)
### [tf.stop_gradient](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/stop_gradient)
``` python
tf.stop_gradient(
input,
......@@ -15,4 +14,4 @@ TensorFlow中，使用`stop_gradient`表示该tensor不需要进行bp。而在Pa
## 代码示例
```python
# 将tensor t设置成不需要bp
t.stop_gradient = True
t.stop_gradient = True
\ No newline at end of file
tensorflow2fluid/doc/tf.while_loop.md
浏览文件 @ ede3f485
## tf.while_loop
### [tf.while_loop](https://www.tensorflow.org/api_docs/python/tf/while_loop)
### [tf.while_loop](https://www.tensorflow.org/versions/r1.13/api_docs/python/tf/while_loop)
```python
tf.while_loop(
......@@ -53,4 +53,4 @@ with while_op.block():
i = layers.increment(x=i, in_place=True)
# 更新条件状态
layers.less_than(x=i, y=limit, cond=cond)
```
```
\ No newline at end of file
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或