Skip to content

P
Paddle

机器未来 / Paddle
与 Fork 源项目一致

Fork自 PaddlePaddle / Paddle

通知 1
Star 1
Fork 0
P
Paddle
“ebb92b7830e269a79de9602176b37d326c78fdf7”上不存在“...multi_language_interface/01.inference_implementation.html”
提交 02a44c38 编写于 作者: T Travis CI
浏览文件
操作

Deploy to GitHub Pages: 32b10d3b

上级 5862ec7b
隐藏空白更改
内联 并排
Showing with 152 addition and 142 deletion
develop/doc/_sources/design/reader/README.md.txt
浏览文件 @ 02a44c38
# Python Data Reader Design Doc # Python Data Reader Design Doc
At training and testing time, PaddlePaddle programs need to read data. To ease the users' work to write data reading code, we define that During the training and testing phases, PaddlePaddle programs need to read data. To help the users write code that performs reading input data, we define the following:
- A *reader* is a function that reads data (from file, network, random number generator, etc) and yields data items. - A *reader*: A function that reads data (from file, network, random number generator, etc) and yields the data items.
- A *reader creator* is a function that returns a reader function. - A *reader creator*: A function that returns a reader function.
- A *reader decorator* is a function, which accepts one or more readers, and returns a reader. - A *reader decorator*: A function, which takes in one or more readers, and returns a reader.
- A *batch reader* is a function that reads data (from *reader*, file, network, random number generator, etc) and yields a batch of data items. - A *batch reader*: A function that reads data (from *reader*, file, network, random number generator, etc) and yields a batch of data items.
and provide function which converts reader to batch reader, frequently used reader creators and reader decorators. and also provide a function which can convert a reader to a batch reader, frequently used reader creators and reader decorators.
## Data Reader Interface ## Data Reader Interface
Indeed, *data reader* doesn't have to be a function that reads and yields data items. It can be any function with no parameter that creates a iterable (anything can be used in `for x in iterable`): *Data reader* doesn't have to be a function that reads and yields data items. It can just be any function without any parameters that creates an iterable (anything can be used in `for x in iterable`) as follows:
``` ```
iterable = data_reader() iterable = data_reader()
``` ```
Element produced from the iterable should be a **single** entry of data, **not** a mini batch. That entry of data could be a single item, or a tuple of items. Item should be of [supported type](http://www.paddlepaddle.org/doc/ui/data_provider/pydataprovider2.html?highlight=dense_vector#input-types) (e.g., numpy 1d array of float32, int, list of int) The item produced from the iterable should be a **single** entry of data and **not** a mini batch. The entry of data could be a single item or a tuple of items. Item should be of one of the [supported types](http://www.paddlepaddle.org/doc/ui/data_provider/pydataprovider2.html?highlight=dense_vector#input-types) (e.g., numpy 1d array of float32, int, list of int etc.)
An example implementation for single item data reader creator: An example implementation for single item data reader creator is as follows:
```python ```python
def reader_creator_random_image(width, height): def reader_creator_random_image(width, height):
...@@ -29,7 +29,7 @@ def reader_creator_random_image(width, height): ...@@ -29,7 +29,7 @@ def reader_creator_random_image(width, height):
return reader return reader
``` ```
An example implementation for multiple item data reader creator: An example implementation for multiple item data reader creator is as follows:
```python ```python
def reader_creator_random_image_and_label(width, height, label): def reader_creator_random_image_and_label(width, height, label):
def reader(): def reader():
...@@ -40,9 +40,10 @@ def reader_creator_random_image_and_label(width, height, label): ...@@ -40,9 +40,10 @@ def reader_creator_random_image_and_label(width, height, label):
## Batch Reader Interface ## Batch Reader Interface
*batch reader* can be any function with no parameter that creates a iterable (anything can be used in `for x in iterable`). The output of the iterable should be a batch (list) of data items. Each item inside the list must be a tuple. *Batch reader* can be any function without any parameters that creates an iterable (anything can be used in `for x in iterable`). The output of the iterable should be a batch (list) of data items. Each item inside the list should be a tuple.
Here are some valid outputs:
Here are valid outputs:
```python ```python
# a mini batch of three data items. Each data item consist three columns of data, each of which is 1. # a mini batch of three data items. Each data item consist three columns of data, each of which is 1.
[(1, 1, 1), [(1, 1, 1),
...@@ -58,20 +59,22 @@ Here are valid outputs: ...@@ -58,20 +59,22 @@ Here are valid outputs:
Please note that each item inside the list must be a tuple, below is an invalid output: Please note that each item inside the list must be a tuple, below is an invalid output:
```python ```python
# wrong, [1,1,1] needs to be inside a tuple: ([1,1,1],). # wrong, [1,1,1] needs to be inside a tuple: ([1,1,1],).
# Otherwise it's ambiguous whether [1,1,1] means a single column of data [1, 1, 1], # Otherwise it is ambiguous whether [1,1,1] means a single column of data [1, 1, 1],
# or three column of datas, each of which is 1. # or three columns of data, each of which is 1.
[[1,1,1], [[1,1,1],
[2,2,2], [2,2,2],
[3,3,3]] [3,3,3]]
``` ```
It's easy to convert from reader to batch reader: It is easy to convert from a reader to a batch reader:
```python ```python
mnist_train = paddle.dataset.mnist.train() mnist_train = paddle.dataset.mnist.train()
mnist_train_batch_reader = paddle.batch(mnist_train, 128) mnist_train_batch_reader = paddle.batch(mnist_train, 128)
``` ```
Also easy to create custom batch reader: It is also straight forward to create a custom batch reader:
```python ```python
def custom_batch_reader(): def custom_batch_reader():
while True: while True:
...@@ -85,7 +88,8 @@ mnist_random_image_batch_reader = custom_batch_reader ...@@ -85,7 +88,8 @@ mnist_random_image_batch_reader = custom_batch_reader
## Usage ## Usage
batch reader, mapping from item(s) read to data layer, batch size and number of total pass will be passed into `paddle.train`: Following is how we can use the reader with PaddlePaddle:
The batch reader, a mapping from item(s) to data layer, the batch size and the number of total passes will be passed into `paddle.train` as follows:
```python ```python
# two data layer is created: # two data layer is created:
...@@ -99,13 +103,13 @@ paddle.train(batch_reader, {"image":0, "label":1}, 128, 10, ...) ...@@ -99,13 +103,13 @@ paddle.train(batch_reader, {"image":0, "label":1}, 128, 10, ...)
## Data Reader Decorator ## Data Reader Decorator
*Data reader decorator* takes a single or multiple data reader, returns a new data reader. It is similar to a [python decorator](https://wiki.python.org/moin/PythonDecorators), but it does not use `@` syntax. The *Data reader decorator* takes in a single reader or multiple data readers and returns a new data reader. It is similar to a [python decorator](https://wiki.python.org/moin/PythonDecorators), but it does not use `@` in the syntax.
Since we have a strict interface for data readers (no parameter, return a single data item). Data reader can be used flexiable via data reader decorators. Following are a few examples: Since we have a strict interface for data readers (no parameters and return a single data item), a data reader can be used in a flexible way using data reader decorators. Following are a few examples:
### Prefetch Data ### Prefetch Data
Since reading data may take time and training can not proceed without data. It is generally a good idea to prefetch data. Since reading data may take some time and training can not proceed without data, it is generally a good idea to prefetch the data.
Use `paddle.reader.buffered` to prefetch data: Use `paddle.reader.buffered` to prefetch data:
...@@ -117,9 +121,9 @@ buffered_reader = paddle.reader.buffered(paddle.dataset.mnist.train(), 100) ...@@ -117,9 +121,9 @@ buffered_reader = paddle.reader.buffered(paddle.dataset.mnist.train(), 100)
### Compose Multiple Data Readers ### Compose Multiple Data Readers
For example, we want to use a source of real images (reusing mnist dataset), and a source of random images as input for [Generative Adversarial Networks](https://arxiv.org/abs/1406.2661). For example, if we want to use a source of real images (say reusing mnist dataset), and a source of random images as input for [Generative Adversarial Networks](https://arxiv.org/abs/1406.2661).
We can do: We can do the following :
```python ```python
def reader_creator_random_image(width, height): def reader_creator_random_image(width, height):
...@@ -139,13 +143,13 @@ false_reader = reader_creator_bool(False) ...@@ -139,13 +143,13 @@ false_reader = reader_creator_bool(False)
reader = paddle.reader.compose(paddle.dataset.mnist.train(), data_reader_creator_random_image(20, 20), true_reader, false_reader) reader = paddle.reader.compose(paddle.dataset.mnist.train(), data_reader_creator_random_image(20, 20), true_reader, false_reader)
# Skipped 1 because paddle.dataset.mnist.train() produces two items per data entry. # Skipped 1 because paddle.dataset.mnist.train() produces two items per data entry.
# And we don't care second item at this time. # And we don't care about the second item at this time.
paddle.train(paddle.batch(reader, 128), {"true_image":0, "fake_image": 2, "true_label": 3, "false_label": 4}, ...) paddle.train(paddle.batch(reader, 128), {"true_image":0, "fake_image": 2, "true_label": 3, "false_label": 4}, ...)
``` ```
### Shuffle ### Shuffle
Given shuffle buffer size `n`, `paddle.reader.shuffle` will return a data reader that buffers `n` data entries and shuffle them before a data entry is read. Given the shuffle buffer size `n`, `paddle.reader.shuffle` returns a data reader that buffers `n` data entries and shuffles them before a data entry is read.
Example: Example:
```python ```python
...@@ -154,21 +158,21 @@ reader = paddle.reader.shuffle(paddle.dataset.mnist.train(), 512) ...@@ -154,21 +158,21 @@ reader = paddle.reader.shuffle(paddle.dataset.mnist.train(), 512)
## Q & A ## Q & A
### Why reader return only a single entry, but not a mini batch? ### Why does a reader return only a single entry, and not a mini batch?
Always returning a single entry make reusing existing data readers much easier (e.g., if existing reader return not a single entry but 3 entries, training code will be more complex because it need to handle cases like batch size 2). Returning a single entry makes reusing existing data readers much easier (for example, if an existing reader returns 3 entries instead if a single entry, the training code will be more complicated because it need to handle cases like a batch size 2).
We provide function `paddle.batch` to turn (single entry) reader into batch reader. We provide a function: `paddle.batch` to turn (a single entry) reader into a batch reader.
### Why do we need batch reader, isn't train take reader and batch_size as arguments sufficient? ### Why do we need a batch reader, isn't is sufficient to give the reader and batch_size as arguments during training ?
In most of the case, train taking reader and batch_size as arguments would be sufficent. However sometimes user want to customize order of data entries inside a mini batch. Or even change batch size dynamically. In most of the cases, it would be sufficient to give the reader and batch_size as arguments to the train method. However sometimes the user wants to customize the order of data entries inside a mini batch, or even change the batch size dynamically. For these cases using a batch reader is very efficient and helpful.
### Why use a dictionary but not a list to provide mapping? ### Why use a dictionary instead of a list to provide mapping?
We decided to use dictionary (`{"image":0, "label":1}`) instead of list (`["image", "label"]`) is because that user can easily resue item (e.g., using `{"image_a":0, "image_b":0, "label":1}`) or skip item (e.g., using `{"image_a":0, "label":2}`). Using a dictionary (`{"image":0, "label":1}`) instead of a list (`["image", "label"]`) gives the advantage that the user can easily reuse the items (e.g., using `{"image_a":0, "image_b":0, "label":1}`) or even skip an item (e.g., using `{"image_a":0, "label":2}`).
### How to create custom data reader creator ### How to create a custom data reader creator ?
```python ```python
def image_reader_creator(image_path, label_path, n): def image_reader_creator(image_path, label_path, n):
...@@ -192,7 +196,7 @@ paddle.train(paddle.batch(reader, 128), {"image":0, "label":1}, ...) ...@@ -192,7 +196,7 @@ paddle.train(paddle.batch(reader, 128), {"image":0, "label":1}, ...)
### How is `paddle.train` implemented ### How is `paddle.train` implemented
An example implementation of paddle.train could be: An example implementation of paddle.train is:
```python ```python
def train(batch_reader, mapping, batch_size, total_pass): def train(batch_reader, mapping, batch_size, total_pass):
......
develop/doc/design/reader/README.html
浏览文件 @ 02a44c38
...@@ -190,22 +190,22 @@ ...@@ -190,22 +190,22 @@
<div class="section" id="python-data-reader-design-doc"> <div class="section" id="python-data-reader-design-doc">
<span id="python-data-reader-design-doc"></span><h1>Python Data Reader Design Doc<a class="headerlink" href="#python-data-reader-design-doc" title="Permalink to this headline"></a></h1> <span id="python-data-reader-design-doc"></span><h1>Python Data Reader Design Doc<a class="headerlink" href="#python-data-reader-design-doc" title="Permalink to this headline"></a></h1>
<p>At training and testing time, PaddlePaddle programs need to read data. To ease the users&#8217; work to write data reading code, we define that</p> <p>During the training and testing phases, PaddlePaddle programs need to read data. To help the users write code that performs reading input data, we define the following:</p>
<ul class="simple"> <ul class="simple">
<li>A <em>reader</em> is a function that reads data (from file, network, random number generator, etc) and yields data items.</li> <li>A <em>reader</em>: A function that reads data (from file, network, random number generator, etc) and yields the data items.</li>
<li>A <em>reader creator</em> is a function that returns a reader function.</li> <li>A <em>reader creator</em>: A function that returns a reader function.</li>
<li>A <em>reader decorator</em> is a function, which accepts one or more readers, and returns a reader.</li> <li>A <em>reader decorator</em>: A function, which takes in one or more readers, and returns a reader.</li>
<li>A <em>batch reader</em> is a function that reads data (from <em>reader</em>, file, network, random number generator, etc) and yields a batch of data items.</li> <li>A <em>batch reader</em>: A function that reads data (from <em>reader</em>, file, network, random number generator, etc) and yields a batch of data items.</li>
</ul> </ul>
<p>and provide function which converts reader to batch reader, frequently used reader creators and reader decorators.</p> <p>and also provide a function which can convert a reader to a batch reader, frequently used reader creators and reader decorators.</p>
<div class="section" id="data-reader-interface"> <div class="section" id="data-reader-interface">
<span id="data-reader-interface"></span><h2>Data Reader Interface<a class="headerlink" href="#data-reader-interface" title="Permalink to this headline"></a></h2> <span id="data-reader-interface"></span><h2>Data Reader Interface<a class="headerlink" href="#data-reader-interface" title="Permalink to this headline"></a></h2>
<p>Indeed, <em>data reader</em> doesn&#8217;t have to be a function that reads and yields data items. It can be any function with no parameter that creates a iterable (anything can be used in <code class="docutils literal"><span class="pre">for</span> <span class="pre">x</span> <span class="pre">in</span> <span class="pre">iterable</span></code>):</p> <p><em>Data reader</em> doesn&#8217;t have to be a function that reads and yields data items. It can just be any function without any parameters that creates an iterable (anything can be used in <code class="docutils literal"><span class="pre">for</span> <span class="pre">x</span> <span class="pre">in</span> <span class="pre">iterable</span></code>) as follows:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="n">iterable</span> <span class="o">=</span> <span class="n">data_reader</span><span class="p">()</span> <div class="highlight-default"><div class="highlight"><pre><span></span><span class="n">iterable</span> <span class="o">=</span> <span class="n">data_reader</span><span class="p">()</span>
</pre></div> </pre></div>
</div> </div>
<p>Element produced from the iterable should be a <strong>single</strong> entry of data, <strong>not</strong> a mini batch. That entry of data could be a single item, or a tuple of items. Item should be of <a class="reference external" href="http://www.paddlepaddle.org/doc/ui/data_provider/pydataprovider2.html?highlight=dense_vector#input-types">supported type</a> (e.g., numpy 1d array of float32, int, list of int)</p> <p>The item produced from the iterable should be a <strong>single</strong> entry of data and <strong>not</strong> a mini batch. The entry of data could be a single item or a tuple of items. Item should be of one of the <a class="reference external" href="http://www.paddlepaddle.org/doc/ui/data_provider/pydataprovider2.html?highlight=dense_vector#input-types">supported types</a> (e.g., numpy 1d array of float32, int, list of int etc.)</p>
<p>An example implementation for single item data reader creator:</p> <p>An example implementation for single item data reader creator is as follows:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">reader</span><span class="p">():</span> <span class="k">def</span> <span class="nf">reader</span><span class="p">():</span>
<span class="k">while</span> <span class="bp">True</span><span class="p">:</span> <span class="k">while</span> <span class="bp">True</span><span class="p">:</span>
...@@ -213,7 +213,7 @@ ...@@ -213,7 +213,7 @@
<span class="k">return</span> <span class="n">reader</span> <span class="k">return</span> <span class="n">reader</span>
</pre></div> </pre></div>
</div> </div>
<p>An example implementation for multiple item data reader creator:</p> <p>An example implementation for multiple item data reader creator is as follows:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image_and_label</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">label</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image_and_label</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">label</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">reader</span><span class="p">():</span> <span class="k">def</span> <span class="nf">reader</span><span class="p">():</span>
<span class="k">while</span> <span class="bp">True</span><span class="p">:</span> <span class="k">while</span> <span class="bp">True</span><span class="p">:</span>
...@@ -224,8 +224,8 @@ ...@@ -224,8 +224,8 @@
</div> </div>
<div class="section" id="batch-reader-interface"> <div class="section" id="batch-reader-interface">
<span id="batch-reader-interface"></span><h2>Batch Reader Interface<a class="headerlink" href="#batch-reader-interface" title="Permalink to this headline"></a></h2> <span id="batch-reader-interface"></span><h2>Batch Reader Interface<a class="headerlink" href="#batch-reader-interface" title="Permalink to this headline"></a></h2>
<p><em>batch reader</em> can be any function with no parameter that creates a iterable (anything can be used in <code class="docutils literal"><span class="pre">for</span> <span class="pre">x</span> <span class="pre">in</span> <span class="pre">iterable</span></code>). The output of the iterable should be a batch (list) of data items. Each item inside the list must be a tuple.</p> <p><em>Batch reader</em> can be any function without any parameters that creates an iterable (anything can be used in <code class="docutils literal"><span class="pre">for</span> <span class="pre">x</span> <span class="pre">in</span> <span class="pre">iterable</span></code>). The output of the iterable should be a batch (list) of data items. Each item inside the list should be a tuple.</p>
<p>Here are valid outputs:</p> <p>Here are some valid outputs:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># a mini batch of three data items. Each data item consist three columns of data, each of which is 1.</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># a mini batch of three data items. Each data item consist three columns of data, each of which is 1.</span>
<span class="p">[(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span> <span class="p">[(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
<span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">),</span> <span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">),</span>
...@@ -239,19 +239,19 @@ ...@@ -239,19 +239,19 @@
</div> </div>
<p>Please note that each item inside the list must be a tuple, below is an invalid output:</p> <p>Please note that each item inside the list must be a tuple, below is an invalid output:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span> <span class="c1"># wrong, [1,1,1] needs to be inside a tuple: ([1,1,1],).</span> <div class="highlight-python"><div class="highlight"><pre><span></span> <span class="c1"># wrong, [1,1,1] needs to be inside a tuple: ([1,1,1],).</span>
<span class="c1"># Otherwise it&#39;s ambiguous whether [1,1,1] means a single column of data [1, 1, 1],</span> <span class="c1"># Otherwise it is ambiguous whether [1,1,1] means a single column of data [1, 1, 1],</span>
<span class="c1"># or three column of datas, each of which is 1.</span> <span class="c1"># or three columns of data, each of which is 1.</span>
<span class="p">[[</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">],</span> <span class="p">[[</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">],</span>
<span class="p">[</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">],</span> <span class="p">[</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">],</span>
<span class="p">[</span><span class="mi">3</span><span class="p">,</span><span class="mi">3</span><span class="p">,</span><span class="mi">3</span><span class="p">]]</span> <span class="p">[</span><span class="mi">3</span><span class="p">,</span><span class="mi">3</span><span class="p">,</span><span class="mi">3</span><span class="p">]]</span>
</pre></div> </pre></div>
</div> </div>
<p>It&#8217;s easy to convert from reader to batch reader:</p> <p>It is easy to convert from a reader to a batch reader:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">mnist_train</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">()</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">mnist_train</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">()</span>
<span class="n">mnist_train_batch_reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">batch</span><span class="p">(</span><span class="n">mnist_train</span><span class="p">,</span> <span class="mi">128</span><span class="p">)</span> <span class="n">mnist_train_batch_reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">batch</span><span class="p">(</span><span class="n">mnist_train</span><span class="p">,</span> <span class="mi">128</span><span class="p">)</span>
</pre></div> </pre></div>
</div> </div>
<p>Also easy to create custom batch reader:</p> <p>It is also straight forward to create a custom batch reader:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">custom_batch_reader</span><span class="p">():</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">custom_batch_reader</span><span class="p">():</span>
<span class="k">while</span> <span class="bp">True</span><span class="p">:</span> <span class="k">while</span> <span class="bp">True</span><span class="p">:</span>
<span class="n">batch</span> <span class="o">=</span> <span class="p">[]</span> <span class="n">batch</span> <span class="o">=</span> <span class="p">[]</span>
...@@ -265,7 +265,8 @@ ...@@ -265,7 +265,8 @@
</div> </div>
<div class="section" id="usage"> <div class="section" id="usage">
<span id="usage"></span><h2>Usage<a class="headerlink" href="#usage" title="Permalink to this headline"></a></h2> <span id="usage"></span><h2>Usage<a class="headerlink" href="#usage" title="Permalink to this headline"></a></h2>
<p>batch reader, mapping from item(s) read to data layer, batch size and number of total pass will be passed into <code class="docutils literal"><span class="pre">paddle.train</span></code>:</p> <p>Following is how we can use the reader with PaddlePaddle:
The batch reader, a mapping from item(s) to data layer, the batch size and the number of total passes will be passed into <code class="docutils literal"><span class="pre">paddle.train</span></code> as follows:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># two data layer is created:</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># two data layer is created:</span>
<span class="n">image_layer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">layer</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="s2">&quot;image&quot;</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span> <span class="n">image_layer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">layer</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="s2">&quot;image&quot;</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span>
<span class="n">label_layer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">layer</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="s2">&quot;label&quot;</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span> <span class="n">label_layer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">layer</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="s2">&quot;label&quot;</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span>
...@@ -278,11 +279,11 @@ ...@@ -278,11 +279,11 @@
</div> </div>
<div class="section" id="data-reader-decorator"> <div class="section" id="data-reader-decorator">
<span id="data-reader-decorator"></span><h2>Data Reader Decorator<a class="headerlink" href="#data-reader-decorator" title="Permalink to this headline"></a></h2> <span id="data-reader-decorator"></span><h2>Data Reader Decorator<a class="headerlink" href="#data-reader-decorator" title="Permalink to this headline"></a></h2>
<p><em>Data reader decorator</em> takes a single or multiple data reader, returns a new data reader. It is similar to a <a class="reference external" href="https://wiki.python.org/moin/PythonDecorators">python decorator</a>, but it does not use <code class="docutils literal"><span class="pre">&#64;</span></code> syntax.</p> <p>The <em>Data reader decorator</em> takes in a single reader or multiple data readers and returns a new data reader. It is similar to a <a class="reference external" href="https://wiki.python.org/moin/PythonDecorators">python decorator</a>, but it does not use <code class="docutils literal"><span class="pre">&#64;</span></code> in the syntax.</p>
<p>Since we have a strict interface for data readers (no parameter, return a single data item). Data reader can be used flexiable via data reader decorators. Following are a few examples:</p> <p>Since we have a strict interface for data readers (no parameters and return a single data item), a data reader can be used in a flexible way using data reader decorators. Following are a few examples:</p>
<div class="section" id="prefetch-data"> <div class="section" id="prefetch-data">
<span id="prefetch-data"></span><h3>Prefetch Data<a class="headerlink" href="#prefetch-data" title="Permalink to this headline"></a></h3> <span id="prefetch-data"></span><h3>Prefetch Data<a class="headerlink" href="#prefetch-data" title="Permalink to this headline"></a></h3>
<p>Since reading data may take time and training can not proceed without data. It is generally a good idea to prefetch data.</p> <p>Since reading data may take some time and training can not proceed without data, it is generally a good idea to prefetch the data.</p>
<p>Use <code class="docutils literal"><span class="pre">paddle.reader.buffered</span></code> to prefetch data:</p> <p>Use <code class="docutils literal"><span class="pre">paddle.reader.buffered</span></code> to prefetch data:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">buffered_reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">buffered</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="mi">100</span><span class="p">)</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">buffered_reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">buffered</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="mi">100</span><span class="p">)</span>
</pre></div> </pre></div>
...@@ -291,8 +292,8 @@ ...@@ -291,8 +292,8 @@
</div> </div>
<div class="section" id="compose-multiple-data-readers"> <div class="section" id="compose-multiple-data-readers">
<span id="compose-multiple-data-readers"></span><h3>Compose Multiple Data Readers<a class="headerlink" href="#compose-multiple-data-readers" title="Permalink to this headline"></a></h3> <span id="compose-multiple-data-readers"></span><h3>Compose Multiple Data Readers<a class="headerlink" href="#compose-multiple-data-readers" title="Permalink to this headline"></a></h3>
<p>For example, we want to use a source of real images (reusing mnist dataset), and a source of random images as input for <a class="reference external" href="https://arxiv.org/abs/1406.2661">Generative Adversarial Networks</a>.</p> <p>For example, if we want to use a source of real images (say reusing mnist dataset), and a source of random images as input for <a class="reference external" href="https://arxiv.org/abs/1406.2661">Generative Adversarial Networks</a>.</p>
<p>We can do:</p> <p>We can do the following :</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">reader</span><span class="p">():</span> <span class="k">def</span> <span class="nf">reader</span><span class="p">():</span>
<span class="k">while</span> <span class="bp">True</span><span class="p">:</span> <span class="k">while</span> <span class="bp">True</span><span class="p">:</span>
...@@ -310,14 +311,14 @@ ...@@ -310,14 +311,14 @@
<span class="n">reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">compose</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="n">data_reader_creator_random_image</span><span class="p">(</span><span class="mi">20</span><span class="p">,</span> <span class="mi">20</span><span class="p">),</span> <span class="n">true_reader</span><span class="p">,</span> <span class="n">false_reader</span><span class="p">)</span> <span class="n">reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">compose</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="n">data_reader_creator_random_image</span><span class="p">(</span><span class="mi">20</span><span class="p">,</span> <span class="mi">20</span><span class="p">),</span> <span class="n">true_reader</span><span class="p">,</span> <span class="n">false_reader</span><span class="p">)</span>
<span class="c1"># Skipped 1 because paddle.dataset.mnist.train() produces two items per data entry.</span> <span class="c1"># Skipped 1 because paddle.dataset.mnist.train() produces two items per data entry.</span>
<span class="c1"># And we don&#39;t care second item at this time.</span> <span class="c1"># And we don&#39;t care about the second item at this time.</span>
<span class="n">paddle</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">batch</span><span class="p">(</span><span class="n">reader</span><span class="p">,</span> <span class="mi">128</span><span class="p">),</span> <span class="p">{</span><span class="s2">&quot;true_image&quot;</span><span class="p">:</span><span class="mi">0</span><span class="p">,</span> <span class="s2">&quot;fake_image&quot;</span><span class="p">:</span> <span class="mi">2</span><span class="p">,</span> <span class="s2">&quot;true_label&quot;</span><span class="p">:</span> <span class="mi">3</span><span class="p">,</span> <span class="s2">&quot;false_label&quot;</span><span class="p">:</span> <span class="mi">4</span><span class="p">},</span> <span class="o">...</span><span class="p">)</span> <span class="n">paddle</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">batch</span><span class="p">(</span><span class="n">reader</span><span class="p">,</span> <span class="mi">128</span><span class="p">),</span> <span class="p">{</span><span class="s2">&quot;true_image&quot;</span><span class="p">:</span><span class="mi">0</span><span class="p">,</span> <span class="s2">&quot;fake_image&quot;</span><span class="p">:</span> <span class="mi">2</span><span class="p">,</span> <span class="s2">&quot;true_label&quot;</span><span class="p">:</span> <span class="mi">3</span><span class="p">,</span> <span class="s2">&quot;false_label&quot;</span><span class="p">:</span> <span class="mi">4</span><span class="p">},</span> <span class="o">...</span><span class="p">)</span>
</pre></div> </pre></div>
</div> </div>
</div> </div>
<div class="section" id="shuffle"> <div class="section" id="shuffle">
<span id="shuffle"></span><h3>Shuffle<a class="headerlink" href="#shuffle" title="Permalink to this headline"></a></h3> <span id="shuffle"></span><h3>Shuffle<a class="headerlink" href="#shuffle" title="Permalink to this headline"></a></h3>
<p>Given shuffle buffer size <code class="docutils literal"><span class="pre">n</span></code>, <code class="docutils literal"><span class="pre">paddle.reader.shuffle</span></code> will return a data reader that buffers <code class="docutils literal"><span class="pre">n</span></code> data entries and shuffle them before a data entry is read.</p> <p>Given the shuffle buffer size <code class="docutils literal"><span class="pre">n</span></code>, <code class="docutils literal"><span class="pre">paddle.reader.shuffle</span></code> returns a data reader that buffers <code class="docutils literal"><span class="pre">n</span></code> data entries and shuffles them before a data entry is read.</p>
<p>Example:</p> <p>Example:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">shuffle</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="mi">512</span><span class="p">)</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">shuffle</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="mi">512</span><span class="p">)</span>
</pre></div> </pre></div>
...@@ -326,21 +327,21 @@ ...@@ -326,21 +327,21 @@
</div> </div>
<div class="section" id="q-a"> <div class="section" id="q-a">
<span id="q-a"></span><h2>Q &amp; A<a class="headerlink" href="#q-a" title="Permalink to this headline"></a></h2> <span id="q-a"></span><h2>Q &amp; A<a class="headerlink" href="#q-a" title="Permalink to this headline"></a></h2>
<div class="section" id="why-reader-return-only-a-single-entry-but-not-a-mini-batch"> <div class="section" id="why-does-a-reader-return-only-a-single-entry-and-not-a-mini-batch">
<span id="why-reader-return-only-a-single-entry-but-not-a-mini-batch"></span><h3>Why reader return only a single entry, but not a mini batch?<a class="headerlink" href="#why-reader-return-only-a-single-entry-but-not-a-mini-batch" title="Permalink to this headline"></a></h3> <span id="why-does-a-reader-return-only-a-single-entry-and-not-a-mini-batch"></span><h3>Why does a reader return only a single entry, and not a mini batch?<a class="headerlink" href="#why-does-a-reader-return-only-a-single-entry-and-not-a-mini-batch" title="Permalink to this headline"></a></h3>
<p>Always returning a single entry make reusing existing data readers much easier (e.g., if existing reader return not a single entry but 3 entries, training code will be more complex because it need to handle cases like batch size 2).</p> <p>Returning a single entry makes reusing existing data readers much easier (for example, if an existing reader returns 3 entries instead if a single entry, the training code will be more complicated because it need to handle cases like a batch size 2).</p>
<p>We provide function <code class="docutils literal"><span class="pre">paddle.batch</span></code> to turn (single entry) reader into batch reader.</p> <p>We provide a function: <code class="docutils literal"><span class="pre">paddle.batch</span></code> to turn (a single entry) reader into a batch reader.</p>
</div> </div>
<div class="section" id="why-do-we-need-batch-reader-isn-t-train-take-reader-and-batch-size-as-arguments-sufficient"> <div class="section" id="why-do-we-need-a-batch-reader-isn-t-is-sufficient-to-give-the-reader-and-batch-size-as-arguments-during-training">
<span id="why-do-we-need-batch-reader-isn-t-train-take-reader-and-batch-size-as-arguments-sufficient"></span><h3>Why do we need batch reader, isn&#8217;t train take reader and batch_size as arguments sufficient?<a class="headerlink" href="#why-do-we-need-batch-reader-isn-t-train-take-reader-and-batch-size-as-arguments-sufficient" title="Permalink to this headline"></a></h3> <span id="why-do-we-need-a-batch-reader-isn-t-is-sufficient-to-give-the-reader-and-batch-size-as-arguments-during-training"></span><h3>Why do we need a batch reader, isn&#8217;t is sufficient to give the reader and batch_size as arguments during training ?<a class="headerlink" href="#why-do-we-need-a-batch-reader-isn-t-is-sufficient-to-give-the-reader-and-batch-size-as-arguments-during-training" title="Permalink to this headline"></a></h3>
<p>In most of the case, train taking reader and batch_size as arguments would be sufficent. However sometimes user want to customize order of data entries inside a mini batch. Or even change batch size dynamically.</p> <p>In most of the cases, it would be sufficient to give the reader and batch_size as arguments to the train method. However sometimes the user wants to customize the order of data entries inside a mini batch, or even change the batch size dynamically. For these cases using a batch reader is very efficient and helpful.</p>
</div> </div>
<div class="section" id="why-use-a-dictionary-but-not-a-list-to-provide-mapping"> <div class="section" id="why-use-a-dictionary-instead-of-a-list-to-provide-mapping">
<span id="why-use-a-dictionary-but-not-a-list-to-provide-mapping"></span><h3>Why use a dictionary but not a list to provide mapping?<a class="headerlink" href="#why-use-a-dictionary-but-not-a-list-to-provide-mapping" title="Permalink to this headline"></a></h3> <span id="why-use-a-dictionary-instead-of-a-list-to-provide-mapping"></span><h3>Why use a dictionary instead of a list to provide mapping?<a class="headerlink" href="#why-use-a-dictionary-instead-of-a-list-to-provide-mapping" title="Permalink to this headline"></a></h3>
<p>We decided to use dictionary (<code class="docutils literal"><span class="pre">{&quot;image&quot;:0,</span> <span class="pre">&quot;label&quot;:1}</span></code>) instead of list (<code class="docutils literal"><span class="pre">[&quot;image&quot;,</span> <span class="pre">&quot;label&quot;]</span></code>) is because that user can easily resue item (e.g., using <code class="docutils literal"><span class="pre">{&quot;image_a&quot;:0,</span> <span class="pre">&quot;image_b&quot;:0,</span> <span class="pre">&quot;label&quot;:1}</span></code>) or skip item (e.g., using <code class="docutils literal"><span class="pre">{&quot;image_a&quot;:0,</span> <span class="pre">&quot;label&quot;:2}</span></code>).</p> <p>Using a dictionary (<code class="docutils literal"><span class="pre">{&quot;image&quot;:0,</span> <span class="pre">&quot;label&quot;:1}</span></code>) instead of a list (<code class="docutils literal"><span class="pre">[&quot;image&quot;,</span> <span class="pre">&quot;label&quot;]</span></code>) gives the advantage that the user can easily reuse the items (e.g., using <code class="docutils literal"><span class="pre">{&quot;image_a&quot;:0,</span> <span class="pre">&quot;image_b&quot;:0,</span> <span class="pre">&quot;label&quot;:1}</span></code>) or even skip an item (e.g., using <code class="docutils literal"><span class="pre">{&quot;image_a&quot;:0,</span> <span class="pre">&quot;label&quot;:2}</span></code>).</p>
</div> </div>
<div class="section" id="how-to-create-custom-data-reader-creator"> <div class="section" id="how-to-create-a-custom-data-reader-creator">
<span id="how-to-create-custom-data-reader-creator"></span><h3>How to create custom data reader creator<a class="headerlink" href="#how-to-create-custom-data-reader-creator" title="Permalink to this headline"></a></h3> <span id="how-to-create-a-custom-data-reader-creator"></span><h3>How to create a custom data reader creator ?<a class="headerlink" href="#how-to-create-a-custom-data-reader-creator" title="Permalink to this headline"></a></h3>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">image_reader_creator</span><span class="p">(</span><span class="n">image_path</span><span class="p">,</span> <span class="n">label_path</span><span class="p">,</span> <span class="n">n</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">image_reader_creator</span><span class="p">(</span><span class="n">image_path</span><span class="p">,</span> <span class="n">label_path</span><span class="p">,</span> <span class="n">n</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">reader</span><span class="p">():</span> <span class="k">def</span> <span class="nf">reader</span><span class="p">():</span>
<span class="n">f</span> <span class="o">=</span> <span class="nb">open</span><span class="p">(</span><span class="n">image_path</span><span class="p">)</span> <span class="n">f</span> <span class="o">=</span> <span class="nb">open</span><span class="p">(</span><span class="n">image_path</span><span class="p">)</span>
...@@ -363,7 +364,7 @@ ...@@ -363,7 +364,7 @@
</div> </div>
<div class="section" id="how-is-paddle-train-implemented"> <div class="section" id="how-is-paddle-train-implemented">
<span id="how-is-paddle-train-implemented"></span><h3>How is <code class="docutils literal"><span class="pre">paddle.train</span></code> implemented<a class="headerlink" href="#how-is-paddle-train-implemented" title="Permalink to this headline"></a></h3> <span id="how-is-paddle-train-implemented"></span><h3>How is <code class="docutils literal"><span class="pre">paddle.train</span></code> implemented<a class="headerlink" href="#how-is-paddle-train-implemented" title="Permalink to this headline"></a></h3>
<p>An example implementation of paddle.train could be:</p> <p>An example implementation of paddle.train is:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">train</span><span class="p">(</span><span class="n">batch_reader</span><span class="p">,</span> <span class="n">mapping</span><span class="p">,</span> <span class="n">batch_size</span><span class="p">,</span> <span class="n">total_pass</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">train</span><span class="p">(</span><span class="n">batch_reader</span><span class="p">,</span> <span class="n">mapping</span><span class="p">,</span> <span class="n">batch_size</span><span class="p">,</span> <span class="n">total_pass</span><span class="p">):</span>
<span class="k">for</span> <span class="n">pass_idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">total_pass</span><span class="p">):</span> <span class="k">for</span> <span class="n">pass_idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">total_pass</span><span class="p">):</span>
<span class="k">for</span> <span class="n">mini_batch</span> <span class="ow">in</span> <span class="n">batch_reader</span><span class="p">():</span> <span class="c1"># this loop will never end in online learning.</span> <span class="k">for</span> <span class="n">mini_batch</span> <span class="ow">in</span> <span class="n">batch_reader</span><span class="p">():</span> <span class="c1"># this loop will never end in online learning.</span>
......
develop/doc/searchindex.js
浏览文件 @ 02a44c38
因为 它太大了无法显示 source diff 。你可以改为 查看blob。
develop/doc_cn/_sources/design/reader/README.md.txt
浏览文件 @ 02a44c38
# Python Data Reader Design Doc # Python Data Reader Design Doc
At training and testing time, PaddlePaddle programs need to read data. To ease the users' work to write data reading code, we define that During the training and testing phases, PaddlePaddle programs need to read data. To help the users write code that performs reading input data, we define the following:
- A *reader* is a function that reads data (from file, network, random number generator, etc) and yields data items. - A *reader*: A function that reads data (from file, network, random number generator, etc) and yields the data items.
- A *reader creator* is a function that returns a reader function. - A *reader creator*: A function that returns a reader function.
- A *reader decorator* is a function, which accepts one or more readers, and returns a reader. - A *reader decorator*: A function, which takes in one or more readers, and returns a reader.
- A *batch reader* is a function that reads data (from *reader*, file, network, random number generator, etc) and yields a batch of data items. - A *batch reader*: A function that reads data (from *reader*, file, network, random number generator, etc) and yields a batch of data items.
and provide function which converts reader to batch reader, frequently used reader creators and reader decorators. and also provide a function which can convert a reader to a batch reader, frequently used reader creators and reader decorators.
## Data Reader Interface ## Data Reader Interface
Indeed, *data reader* doesn't have to be a function that reads and yields data items. It can be any function with no parameter that creates a iterable (anything can be used in `for x in iterable`): *Data reader* doesn't have to be a function that reads and yields data items. It can just be any function without any parameters that creates an iterable (anything can be used in `for x in iterable`) as follows:
``` ```
iterable = data_reader() iterable = data_reader()
``` ```
Element produced from the iterable should be a **single** entry of data, **not** a mini batch. That entry of data could be a single item, or a tuple of items. Item should be of [supported type](http://www.paddlepaddle.org/doc/ui/data_provider/pydataprovider2.html?highlight=dense_vector#input-types) (e.g., numpy 1d array of float32, int, list of int) The item produced from the iterable should be a **single** entry of data and **not** a mini batch. The entry of data could be a single item or a tuple of items. Item should be of one of the [supported types](http://www.paddlepaddle.org/doc/ui/data_provider/pydataprovider2.html?highlight=dense_vector#input-types) (e.g., numpy 1d array of float32, int, list of int etc.)
An example implementation for single item data reader creator: An example implementation for single item data reader creator is as follows:
```python ```python
def reader_creator_random_image(width, height): def reader_creator_random_image(width, height):
...@@ -29,7 +29,7 @@ def reader_creator_random_image(width, height): ...@@ -29,7 +29,7 @@ def reader_creator_random_image(width, height):
return reader return reader
``` ```
An example implementation for multiple item data reader creator: An example implementation for multiple item data reader creator is as follows:
```python ```python
def reader_creator_random_image_and_label(width, height, label): def reader_creator_random_image_and_label(width, height, label):
def reader(): def reader():
...@@ -40,9 +40,10 @@ def reader_creator_random_image_and_label(width, height, label): ...@@ -40,9 +40,10 @@ def reader_creator_random_image_and_label(width, height, label):
## Batch Reader Interface ## Batch Reader Interface
*batch reader* can be any function with no parameter that creates a iterable (anything can be used in `for x in iterable`). The output of the iterable should be a batch (list) of data items. Each item inside the list must be a tuple. *Batch reader* can be any function without any parameters that creates an iterable (anything can be used in `for x in iterable`). The output of the iterable should be a batch (list) of data items. Each item inside the list should be a tuple.
Here are some valid outputs:
Here are valid outputs:
```python ```python
# a mini batch of three data items. Each data item consist three columns of data, each of which is 1. # a mini batch of three data items. Each data item consist three columns of data, each of which is 1.
[(1, 1, 1), [(1, 1, 1),
...@@ -58,20 +59,22 @@ Here are valid outputs: ...@@ -58,20 +59,22 @@ Here are valid outputs:
Please note that each item inside the list must be a tuple, below is an invalid output: Please note that each item inside the list must be a tuple, below is an invalid output:
```python ```python
# wrong, [1,1,1] needs to be inside a tuple: ([1,1,1],). # wrong, [1,1,1] needs to be inside a tuple: ([1,1,1],).
# Otherwise it's ambiguous whether [1,1,1] means a single column of data [1, 1, 1], # Otherwise it is ambiguous whether [1,1,1] means a single column of data [1, 1, 1],
# or three column of datas, each of which is 1. # or three columns of data, each of which is 1.
[[1,1,1], [[1,1,1],
[2,2,2], [2,2,2],
[3,3,3]] [3,3,3]]
``` ```
It's easy to convert from reader to batch reader: It is easy to convert from a reader to a batch reader:
```python ```python
mnist_train = paddle.dataset.mnist.train() mnist_train = paddle.dataset.mnist.train()
mnist_train_batch_reader = paddle.batch(mnist_train, 128) mnist_train_batch_reader = paddle.batch(mnist_train, 128)
``` ```
Also easy to create custom batch reader: It is also straight forward to create a custom batch reader:
```python ```python
def custom_batch_reader(): def custom_batch_reader():
while True: while True:
...@@ -85,7 +88,8 @@ mnist_random_image_batch_reader = custom_batch_reader ...@@ -85,7 +88,8 @@ mnist_random_image_batch_reader = custom_batch_reader
## Usage ## Usage
batch reader, mapping from item(s) read to data layer, batch size and number of total pass will be passed into `paddle.train`: Following is how we can use the reader with PaddlePaddle:
The batch reader, a mapping from item(s) to data layer, the batch size and the number of total passes will be passed into `paddle.train` as follows:
```python ```python
# two data layer is created: # two data layer is created:
...@@ -99,13 +103,13 @@ paddle.train(batch_reader, {"image":0, "label":1}, 128, 10, ...) ...@@ -99,13 +103,13 @@ paddle.train(batch_reader, {"image":0, "label":1}, 128, 10, ...)
## Data Reader Decorator ## Data Reader Decorator
*Data reader decorator* takes a single or multiple data reader, returns a new data reader. It is similar to a [python decorator](https://wiki.python.org/moin/PythonDecorators), but it does not use `@` syntax. The *Data reader decorator* takes in a single reader or multiple data readers and returns a new data reader. It is similar to a [python decorator](https://wiki.python.org/moin/PythonDecorators), but it does not use `@` in the syntax.
Since we have a strict interface for data readers (no parameter, return a single data item). Data reader can be used flexiable via data reader decorators. Following are a few examples: Since we have a strict interface for data readers (no parameters and return a single data item), a data reader can be used in a flexible way using data reader decorators. Following are a few examples:
### Prefetch Data ### Prefetch Data
Since reading data may take time and training can not proceed without data. It is generally a good idea to prefetch data. Since reading data may take some time and training can not proceed without data, it is generally a good idea to prefetch the data.
Use `paddle.reader.buffered` to prefetch data: Use `paddle.reader.buffered` to prefetch data:
...@@ -117,9 +121,9 @@ buffered_reader = paddle.reader.buffered(paddle.dataset.mnist.train(), 100) ...@@ -117,9 +121,9 @@ buffered_reader = paddle.reader.buffered(paddle.dataset.mnist.train(), 100)
### Compose Multiple Data Readers ### Compose Multiple Data Readers
For example, we want to use a source of real images (reusing mnist dataset), and a source of random images as input for [Generative Adversarial Networks](https://arxiv.org/abs/1406.2661). For example, if we want to use a source of real images (say reusing mnist dataset), and a source of random images as input for [Generative Adversarial Networks](https://arxiv.org/abs/1406.2661).
We can do: We can do the following :
```python ```python
def reader_creator_random_image(width, height): def reader_creator_random_image(width, height):
...@@ -139,13 +143,13 @@ false_reader = reader_creator_bool(False) ...@@ -139,13 +143,13 @@ false_reader = reader_creator_bool(False)
reader = paddle.reader.compose(paddle.dataset.mnist.train(), data_reader_creator_random_image(20, 20), true_reader, false_reader) reader = paddle.reader.compose(paddle.dataset.mnist.train(), data_reader_creator_random_image(20, 20), true_reader, false_reader)
# Skipped 1 because paddle.dataset.mnist.train() produces two items per data entry. # Skipped 1 because paddle.dataset.mnist.train() produces two items per data entry.
# And we don't care second item at this time. # And we don't care about the second item at this time.
paddle.train(paddle.batch(reader, 128), {"true_image":0, "fake_image": 2, "true_label": 3, "false_label": 4}, ...) paddle.train(paddle.batch(reader, 128), {"true_image":0, "fake_image": 2, "true_label": 3, "false_label": 4}, ...)
``` ```
### Shuffle ### Shuffle
Given shuffle buffer size `n`, `paddle.reader.shuffle` will return a data reader that buffers `n` data entries and shuffle them before a data entry is read. Given the shuffle buffer size `n`, `paddle.reader.shuffle` returns a data reader that buffers `n` data entries and shuffles them before a data entry is read.
Example: Example:
```python ```python
...@@ -154,21 +158,21 @@ reader = paddle.reader.shuffle(paddle.dataset.mnist.train(), 512) ...@@ -154,21 +158,21 @@ reader = paddle.reader.shuffle(paddle.dataset.mnist.train(), 512)
## Q & A ## Q & A
### Why reader return only a single entry, but not a mini batch? ### Why does a reader return only a single entry, and not a mini batch?
Always returning a single entry make reusing existing data readers much easier (e.g., if existing reader return not a single entry but 3 entries, training code will be more complex because it need to handle cases like batch size 2). Returning a single entry makes reusing existing data readers much easier (for example, if an existing reader returns 3 entries instead if a single entry, the training code will be more complicated because it need to handle cases like a batch size 2).
We provide function `paddle.batch` to turn (single entry) reader into batch reader. We provide a function: `paddle.batch` to turn (a single entry) reader into a batch reader.
### Why do we need batch reader, isn't train take reader and batch_size as arguments sufficient? ### Why do we need a batch reader, isn't is sufficient to give the reader and batch_size as arguments during training ?
In most of the case, train taking reader and batch_size as arguments would be sufficent. However sometimes user want to customize order of data entries inside a mini batch. Or even change batch size dynamically. In most of the cases, it would be sufficient to give the reader and batch_size as arguments to the train method. However sometimes the user wants to customize the order of data entries inside a mini batch, or even change the batch size dynamically. For these cases using a batch reader is very efficient and helpful.
### Why use a dictionary but not a list to provide mapping? ### Why use a dictionary instead of a list to provide mapping?
We decided to use dictionary (`{"image":0, "label":1}`) instead of list (`["image", "label"]`) is because that user can easily resue item (e.g., using `{"image_a":0, "image_b":0, "label":1}`) or skip item (e.g., using `{"image_a":0, "label":2}`). Using a dictionary (`{"image":0, "label":1}`) instead of a list (`["image", "label"]`) gives the advantage that the user can easily reuse the items (e.g., using `{"image_a":0, "image_b":0, "label":1}`) or even skip an item (e.g., using `{"image_a":0, "label":2}`).
### How to create custom data reader creator ### How to create a custom data reader creator ?
```python ```python
def image_reader_creator(image_path, label_path, n): def image_reader_creator(image_path, label_path, n):
...@@ -192,7 +196,7 @@ paddle.train(paddle.batch(reader, 128), {"image":0, "label":1}, ...) ...@@ -192,7 +196,7 @@ paddle.train(paddle.batch(reader, 128), {"image":0, "label":1}, ...)
### How is `paddle.train` implemented ### How is `paddle.train` implemented
An example implementation of paddle.train could be: An example implementation of paddle.train is:
```python ```python
def train(batch_reader, mapping, batch_size, total_pass): def train(batch_reader, mapping, batch_size, total_pass):
......
develop/doc_cn/design/reader/README.html
浏览文件 @ 02a44c38
...@@ -204,22 +204,22 @@ ...@@ -204,22 +204,22 @@
<div class="section" id="python-data-reader-design-doc"> <div class="section" id="python-data-reader-design-doc">
<span id="python-data-reader-design-doc"></span><h1>Python Data Reader Design Doc<a class="headerlink" href="#python-data-reader-design-doc" title="永久链接至标题"></a></h1> <span id="python-data-reader-design-doc"></span><h1>Python Data Reader Design Doc<a class="headerlink" href="#python-data-reader-design-doc" title="永久链接至标题"></a></h1>
<p>At training and testing time, PaddlePaddle programs need to read data. To ease the users&#8217; work to write data reading code, we define that</p> <p>During the training and testing phases, PaddlePaddle programs need to read data. To help the users write code that performs reading input data, we define the following:</p>
<ul class="simple"> <ul class="simple">
<li>A <em>reader</em> is a function that reads data (from file, network, random number generator, etc) and yields data items.</li> <li>A <em>reader</em>: A function that reads data (from file, network, random number generator, etc) and yields the data items.</li>
<li>A <em>reader creator</em> is a function that returns a reader function.</li> <li>A <em>reader creator</em>: A function that returns a reader function.</li>
<li>A <em>reader decorator</em> is a function, which accepts one or more readers, and returns a reader.</li> <li>A <em>reader decorator</em>: A function, which takes in one or more readers, and returns a reader.</li>
<li>A <em>batch reader</em> is a function that reads data (from <em>reader</em>, file, network, random number generator, etc) and yields a batch of data items.</li> <li>A <em>batch reader</em>: A function that reads data (from <em>reader</em>, file, network, random number generator, etc) and yields a batch of data items.</li>
</ul> </ul>
<p>and provide function which converts reader to batch reader, frequently used reader creators and reader decorators.</p> <p>and also provide a function which can convert a reader to a batch reader, frequently used reader creators and reader decorators.</p>
<div class="section" id="data-reader-interface"> <div class="section" id="data-reader-interface">
<span id="data-reader-interface"></span><h2>Data Reader Interface<a class="headerlink" href="#data-reader-interface" title="永久链接至标题"></a></h2> <span id="data-reader-interface"></span><h2>Data Reader Interface<a class="headerlink" href="#data-reader-interface" title="永久链接至标题"></a></h2>
<p>Indeed, <em>data reader</em> doesn&#8217;t have to be a function that reads and yields data items. It can be any function with no parameter that creates a iterable (anything can be used in <code class="docutils literal"><span class="pre">for</span> <span class="pre">x</span> <span class="pre">in</span> <span class="pre">iterable</span></code>):</p> <p><em>Data reader</em> doesn&#8217;t have to be a function that reads and yields data items. It can just be any function without any parameters that creates an iterable (anything can be used in <code class="docutils literal"><span class="pre">for</span> <span class="pre">x</span> <span class="pre">in</span> <span class="pre">iterable</span></code>) as follows:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="n">iterable</span> <span class="o">=</span> <span class="n">data_reader</span><span class="p">()</span> <div class="highlight-default"><div class="highlight"><pre><span></span><span class="n">iterable</span> <span class="o">=</span> <span class="n">data_reader</span><span class="p">()</span>
</pre></div> </pre></div>
</div> </div>
<p>Element produced from the iterable should be a <strong>single</strong> entry of data, <strong>not</strong> a mini batch. That entry of data could be a single item, or a tuple of items. Item should be of <a class="reference external" href="http://www.paddlepaddle.org/doc/ui/data_provider/pydataprovider2.html?highlight=dense_vector#input-types">supported type</a> (e.g., numpy 1d array of float32, int, list of int)</p> <p>The item produced from the iterable should be a <strong>single</strong> entry of data and <strong>not</strong> a mini batch. The entry of data could be a single item or a tuple of items. Item should be of one of the <a class="reference external" href="http://www.paddlepaddle.org/doc/ui/data_provider/pydataprovider2.html?highlight=dense_vector#input-types">supported types</a> (e.g., numpy 1d array of float32, int, list of int etc.)</p>
<p>An example implementation for single item data reader creator:</p> <p>An example implementation for single item data reader creator is as follows:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">reader</span><span class="p">():</span> <span class="k">def</span> <span class="nf">reader</span><span class="p">():</span>
<span class="k">while</span> <span class="bp">True</span><span class="p">:</span> <span class="k">while</span> <span class="bp">True</span><span class="p">:</span>
...@@ -227,7 +227,7 @@ ...@@ -227,7 +227,7 @@
<span class="k">return</span> <span class="n">reader</span> <span class="k">return</span> <span class="n">reader</span>
</pre></div> </pre></div>
</div> </div>
<p>An example implementation for multiple item data reader creator:</p> <p>An example implementation for multiple item data reader creator is as follows:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image_and_label</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">label</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image_and_label</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">,</span> <span class="n">label</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">reader</span><span class="p">():</span> <span class="k">def</span> <span class="nf">reader</span><span class="p">():</span>
<span class="k">while</span> <span class="bp">True</span><span class="p">:</span> <span class="k">while</span> <span class="bp">True</span><span class="p">:</span>
...@@ -238,8 +238,8 @@ ...@@ -238,8 +238,8 @@
</div> </div>
<div class="section" id="batch-reader-interface"> <div class="section" id="batch-reader-interface">
<span id="batch-reader-interface"></span><h2>Batch Reader Interface<a class="headerlink" href="#batch-reader-interface" title="永久链接至标题"></a></h2> <span id="batch-reader-interface"></span><h2>Batch Reader Interface<a class="headerlink" href="#batch-reader-interface" title="永久链接至标题"></a></h2>
<p><em>batch reader</em> can be any function with no parameter that creates a iterable (anything can be used in <code class="docutils literal"><span class="pre">for</span> <span class="pre">x</span> <span class="pre">in</span> <span class="pre">iterable</span></code>). The output of the iterable should be a batch (list) of data items. Each item inside the list must be a tuple.</p> <p><em>Batch reader</em> can be any function without any parameters that creates an iterable (anything can be used in <code class="docutils literal"><span class="pre">for</span> <span class="pre">x</span> <span class="pre">in</span> <span class="pre">iterable</span></code>). The output of the iterable should be a batch (list) of data items. Each item inside the list should be a tuple.</p>
<p>Here are valid outputs:</p> <p>Here are some valid outputs:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># a mini batch of three data items. Each data item consist three columns of data, each of which is 1.</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># a mini batch of three data items. Each data item consist three columns of data, each of which is 1.</span>
<span class="p">[(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span> <span class="p">[(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
<span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">),</span> <span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">),</span>
...@@ -253,19 +253,19 @@ ...@@ -253,19 +253,19 @@
</div> </div>
<p>Please note that each item inside the list must be a tuple, below is an invalid output:</p> <p>Please note that each item inside the list must be a tuple, below is an invalid output:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span> <span class="c1"># wrong, [1,1,1] needs to be inside a tuple: ([1,1,1],).</span> <div class="highlight-python"><div class="highlight"><pre><span></span> <span class="c1"># wrong, [1,1,1] needs to be inside a tuple: ([1,1,1],).</span>
<span class="c1"># Otherwise it&#39;s ambiguous whether [1,1,1] means a single column of data [1, 1, 1],</span> <span class="c1"># Otherwise it is ambiguous whether [1,1,1] means a single column of data [1, 1, 1],</span>
<span class="c1"># or three column of datas, each of which is 1.</span> <span class="c1"># or three columns of data, each of which is 1.</span>
<span class="p">[[</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">],</span> <span class="p">[[</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">],</span>
<span class="p">[</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">],</span> <span class="p">[</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">],</span>
<span class="p">[</span><span class="mi">3</span><span class="p">,</span><span class="mi">3</span><span class="p">,</span><span class="mi">3</span><span class="p">]]</span> <span class="p">[</span><span class="mi">3</span><span class="p">,</span><span class="mi">3</span><span class="p">,</span><span class="mi">3</span><span class="p">]]</span>
</pre></div> </pre></div>
</div> </div>
<p>It&#8217;s easy to convert from reader to batch reader:</p> <p>It is easy to convert from a reader to a batch reader:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">mnist_train</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">()</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">mnist_train</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">()</span>
<span class="n">mnist_train_batch_reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">batch</span><span class="p">(</span><span class="n">mnist_train</span><span class="p">,</span> <span class="mi">128</span><span class="p">)</span> <span class="n">mnist_train_batch_reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">batch</span><span class="p">(</span><span class="n">mnist_train</span><span class="p">,</span> <span class="mi">128</span><span class="p">)</span>
</pre></div> </pre></div>
</div> </div>
<p>Also easy to create custom batch reader:</p> <p>It is also straight forward to create a custom batch reader:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">custom_batch_reader</span><span class="p">():</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">custom_batch_reader</span><span class="p">():</span>
<span class="k">while</span> <span class="bp">True</span><span class="p">:</span> <span class="k">while</span> <span class="bp">True</span><span class="p">:</span>
<span class="n">batch</span> <span class="o">=</span> <span class="p">[]</span> <span class="n">batch</span> <span class="o">=</span> <span class="p">[]</span>
...@@ -279,7 +279,8 @@ ...@@ -279,7 +279,8 @@
</div> </div>
<div class="section" id="usage"> <div class="section" id="usage">
<span id="usage"></span><h2>Usage<a class="headerlink" href="#usage" title="永久链接至标题"></a></h2> <span id="usage"></span><h2>Usage<a class="headerlink" href="#usage" title="永久链接至标题"></a></h2>
<p>batch reader, mapping from item(s) read to data layer, batch size and number of total pass will be passed into <code class="docutils literal"><span class="pre">paddle.train</span></code>:</p> <p>Following is how we can use the reader with PaddlePaddle:
The batch reader, a mapping from item(s) to data layer, the batch size and the number of total passes will be passed into <code class="docutils literal"><span class="pre">paddle.train</span></code> as follows:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># two data layer is created:</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="c1"># two data layer is created:</span>
<span class="n">image_layer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">layer</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="s2">&quot;image&quot;</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span> <span class="n">image_layer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">layer</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="s2">&quot;image&quot;</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span>
<span class="n">label_layer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">layer</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="s2">&quot;label&quot;</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span> <span class="n">label_layer</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">layer</span><span class="o">.</span><span class="n">data</span><span class="p">(</span><span class="s2">&quot;label&quot;</span><span class="p">,</span> <span class="o">...</span><span class="p">)</span>
...@@ -292,11 +293,11 @@ ...@@ -292,11 +293,11 @@
</div> </div>
<div class="section" id="data-reader-decorator"> <div class="section" id="data-reader-decorator">
<span id="data-reader-decorator"></span><h2>Data Reader Decorator<a class="headerlink" href="#data-reader-decorator" title="永久链接至标题"></a></h2> <span id="data-reader-decorator"></span><h2>Data Reader Decorator<a class="headerlink" href="#data-reader-decorator" title="永久链接至标题"></a></h2>
<p><em>Data reader decorator</em> takes a single or multiple data reader, returns a new data reader. It is similar to a <a class="reference external" href="https://wiki.python.org/moin/PythonDecorators">python decorator</a>, but it does not use <code class="docutils literal"><span class="pre">&#64;</span></code> syntax.</p> <p>The <em>Data reader decorator</em> takes in a single reader or multiple data readers and returns a new data reader. It is similar to a <a class="reference external" href="https://wiki.python.org/moin/PythonDecorators">python decorator</a>, but it does not use <code class="docutils literal"><span class="pre">&#64;</span></code> in the syntax.</p>
<p>Since we have a strict interface for data readers (no parameter, return a single data item). Data reader can be used flexiable via data reader decorators. Following are a few examples:</p> <p>Since we have a strict interface for data readers (no parameters and return a single data item), a data reader can be used in a flexible way using data reader decorators. Following are a few examples:</p>
<div class="section" id="prefetch-data"> <div class="section" id="prefetch-data">
<span id="prefetch-data"></span><h3>Prefetch Data<a class="headerlink" href="#prefetch-data" title="永久链接至标题"></a></h3> <span id="prefetch-data"></span><h3>Prefetch Data<a class="headerlink" href="#prefetch-data" title="永久链接至标题"></a></h3>
<p>Since reading data may take time and training can not proceed without data. It is generally a good idea to prefetch data.</p> <p>Since reading data may take some time and training can not proceed without data, it is generally a good idea to prefetch the data.</p>
<p>Use <code class="docutils literal"><span class="pre">paddle.reader.buffered</span></code> to prefetch data:</p> <p>Use <code class="docutils literal"><span class="pre">paddle.reader.buffered</span></code> to prefetch data:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">buffered_reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">buffered</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="mi">100</span><span class="p">)</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">buffered_reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">buffered</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="mi">100</span><span class="p">)</span>
</pre></div> </pre></div>
...@@ -305,8 +306,8 @@ ...@@ -305,8 +306,8 @@
</div> </div>
<div class="section" id="compose-multiple-data-readers"> <div class="section" id="compose-multiple-data-readers">
<span id="compose-multiple-data-readers"></span><h3>Compose Multiple Data Readers<a class="headerlink" href="#compose-multiple-data-readers" title="永久链接至标题"></a></h3> <span id="compose-multiple-data-readers"></span><h3>Compose Multiple Data Readers<a class="headerlink" href="#compose-multiple-data-readers" title="永久链接至标题"></a></h3>
<p>For example, we want to use a source of real images (reusing mnist dataset), and a source of random images as input for <a class="reference external" href="https://arxiv.org/abs/1406.2661">Generative Adversarial Networks</a>.</p> <p>For example, if we want to use a source of real images (say reusing mnist dataset), and a source of random images as input for <a class="reference external" href="https://arxiv.org/abs/1406.2661">Generative Adversarial Networks</a>.</p>
<p>We can do:</p> <p>We can do the following :</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">reader_creator_random_image</span><span class="p">(</span><span class="n">width</span><span class="p">,</span> <span class="n">height</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">reader</span><span class="p">():</span> <span class="k">def</span> <span class="nf">reader</span><span class="p">():</span>
<span class="k">while</span> <span class="bp">True</span><span class="p">:</span> <span class="k">while</span> <span class="bp">True</span><span class="p">:</span>
...@@ -324,14 +325,14 @@ ...@@ -324,14 +325,14 @@
<span class="n">reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">compose</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="n">data_reader_creator_random_image</span><span class="p">(</span><span class="mi">20</span><span class="p">,</span> <span class="mi">20</span><span class="p">),</span> <span class="n">true_reader</span><span class="p">,</span> <span class="n">false_reader</span><span class="p">)</span> <span class="n">reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">compose</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="n">data_reader_creator_random_image</span><span class="p">(</span><span class="mi">20</span><span class="p">,</span> <span class="mi">20</span><span class="p">),</span> <span class="n">true_reader</span><span class="p">,</span> <span class="n">false_reader</span><span class="p">)</span>
<span class="c1"># Skipped 1 because paddle.dataset.mnist.train() produces two items per data entry.</span> <span class="c1"># Skipped 1 because paddle.dataset.mnist.train() produces two items per data entry.</span>
<span class="c1"># And we don&#39;t care second item at this time.</span> <span class="c1"># And we don&#39;t care about the second item at this time.</span>
<span class="n">paddle</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">batch</span><span class="p">(</span><span class="n">reader</span><span class="p">,</span> <span class="mi">128</span><span class="p">),</span> <span class="p">{</span><span class="s2">&quot;true_image&quot;</span><span class="p">:</span><span class="mi">0</span><span class="p">,</span> <span class="s2">&quot;fake_image&quot;</span><span class="p">:</span> <span class="mi">2</span><span class="p">,</span> <span class="s2">&quot;true_label&quot;</span><span class="p">:</span> <span class="mi">3</span><span class="p">,</span> <span class="s2">&quot;false_label&quot;</span><span class="p">:</span> <span class="mi">4</span><span class="p">},</span> <span class="o">...</span><span class="p">)</span> <span class="n">paddle</span><span class="o">.</span><span class="n">train</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">batch</span><span class="p">(</span><span class="n">reader</span><span class="p">,</span> <span class="mi">128</span><span class="p">),</span> <span class="p">{</span><span class="s2">&quot;true_image&quot;</span><span class="p">:</span><span class="mi">0</span><span class="p">,</span> <span class="s2">&quot;fake_image&quot;</span><span class="p">:</span> <span class="mi">2</span><span class="p">,</span> <span class="s2">&quot;true_label&quot;</span><span class="p">:</span> <span class="mi">3</span><span class="p">,</span> <span class="s2">&quot;false_label&quot;</span><span class="p">:</span> <span class="mi">4</span><span class="p">},</span> <span class="o">...</span><span class="p">)</span>
</pre></div> </pre></div>
</div> </div>
</div> </div>
<div class="section" id="shuffle"> <div class="section" id="shuffle">
<span id="shuffle"></span><h3>Shuffle<a class="headerlink" href="#shuffle" title="永久链接至标题"></a></h3> <span id="shuffle"></span><h3>Shuffle<a class="headerlink" href="#shuffle" title="永久链接至标题"></a></h3>
<p>Given shuffle buffer size <code class="docutils literal"><span class="pre">n</span></code>, <code class="docutils literal"><span class="pre">paddle.reader.shuffle</span></code> will return a data reader that buffers <code class="docutils literal"><span class="pre">n</span></code> data entries and shuffle them before a data entry is read.</p> <p>Given the shuffle buffer size <code class="docutils literal"><span class="pre">n</span></code>, <code class="docutils literal"><span class="pre">paddle.reader.shuffle</span></code> returns a data reader that buffers <code class="docutils literal"><span class="pre">n</span></code> data entries and shuffles them before a data entry is read.</p>
<p>Example:</p> <p>Example:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">shuffle</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="mi">512</span><span class="p">)</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="n">reader</span> <span class="o">=</span> <span class="n">paddle</span><span class="o">.</span><span class="n">reader</span><span class="o">.</span><span class="n">shuffle</span><span class="p">(</span><span class="n">paddle</span><span class="o">.</span><span class="n">dataset</span><span class="o">.</span><span class="n">mnist</span><span class="o">.</span><span class="n">train</span><span class="p">(),</span> <span class="mi">512</span><span class="p">)</span>
</pre></div> </pre></div>
...@@ -340,21 +341,21 @@ ...@@ -340,21 +341,21 @@
</div> </div>
<div class="section" id="q-a"> <div class="section" id="q-a">
<span id="q-a"></span><h2>Q &amp; A<a class="headerlink" href="#q-a" title="永久链接至标题"></a></h2> <span id="q-a"></span><h2>Q &amp; A<a class="headerlink" href="#q-a" title="永久链接至标题"></a></h2>
<div class="section" id="why-reader-return-only-a-single-entry-but-not-a-mini-batch"> <div class="section" id="why-does-a-reader-return-only-a-single-entry-and-not-a-mini-batch">
<span id="why-reader-return-only-a-single-entry-but-not-a-mini-batch"></span><h3>Why reader return only a single entry, but not a mini batch?<a class="headerlink" href="#why-reader-return-only-a-single-entry-but-not-a-mini-batch" title="永久链接至标题"></a></h3> <span id="why-does-a-reader-return-only-a-single-entry-and-not-a-mini-batch"></span><h3>Why does a reader return only a single entry, and not a mini batch?<a class="headerlink" href="#why-does-a-reader-return-only-a-single-entry-and-not-a-mini-batch" title="永久链接至标题"></a></h3>
<p>Always returning a single entry make reusing existing data readers much easier (e.g., if existing reader return not a single entry but 3 entries, training code will be more complex because it need to handle cases like batch size 2).</p> <p>Returning a single entry makes reusing existing data readers much easier (for example, if an existing reader returns 3 entries instead if a single entry, the training code will be more complicated because it need to handle cases like a batch size 2).</p>
<p>We provide function <code class="docutils literal"><span class="pre">paddle.batch</span></code> to turn (single entry) reader into batch reader.</p> <p>We provide a function: <code class="docutils literal"><span class="pre">paddle.batch</span></code> to turn (a single entry) reader into a batch reader.</p>
</div> </div>
<div class="section" id="why-do-we-need-batch-reader-isn-t-train-take-reader-and-batch-size-as-arguments-sufficient"> <div class="section" id="why-do-we-need-a-batch-reader-isn-t-is-sufficient-to-give-the-reader-and-batch-size-as-arguments-during-training">
<span id="why-do-we-need-batch-reader-isn-t-train-take-reader-and-batch-size-as-arguments-sufficient"></span><h3>Why do we need batch reader, isn&#8217;t train take reader and batch_size as arguments sufficient?<a class="headerlink" href="#why-do-we-need-batch-reader-isn-t-train-take-reader-and-batch-size-as-arguments-sufficient" title="永久链接至标题"></a></h3> <span id="why-do-we-need-a-batch-reader-isn-t-is-sufficient-to-give-the-reader-and-batch-size-as-arguments-during-training"></span><h3>Why do we need a batch reader, isn&#8217;t is sufficient to give the reader and batch_size as arguments during training ?<a class="headerlink" href="#why-do-we-need-a-batch-reader-isn-t-is-sufficient-to-give-the-reader-and-batch-size-as-arguments-during-training" title="永久链接至标题"></a></h3>
<p>In most of the case, train taking reader and batch_size as arguments would be sufficent. However sometimes user want to customize order of data entries inside a mini batch. Or even change batch size dynamically.</p> <p>In most of the cases, it would be sufficient to give the reader and batch_size as arguments to the train method. However sometimes the user wants to customize the order of data entries inside a mini batch, or even change the batch size dynamically. For these cases using a batch reader is very efficient and helpful.</p>
</div> </div>
<div class="section" id="why-use-a-dictionary-but-not-a-list-to-provide-mapping"> <div class="section" id="why-use-a-dictionary-instead-of-a-list-to-provide-mapping">
<span id="why-use-a-dictionary-but-not-a-list-to-provide-mapping"></span><h3>Why use a dictionary but not a list to provide mapping?<a class="headerlink" href="#why-use-a-dictionary-but-not-a-list-to-provide-mapping" title="永久链接至标题"></a></h3> <span id="why-use-a-dictionary-instead-of-a-list-to-provide-mapping"></span><h3>Why use a dictionary instead of a list to provide mapping?<a class="headerlink" href="#why-use-a-dictionary-instead-of-a-list-to-provide-mapping" title="永久链接至标题"></a></h3>
<p>We decided to use dictionary (<code class="docutils literal"><span class="pre">{&quot;image&quot;:0,</span> <span class="pre">&quot;label&quot;:1}</span></code>) instead of list (<code class="docutils literal"><span class="pre">[&quot;image&quot;,</span> <span class="pre">&quot;label&quot;]</span></code>) is because that user can easily resue item (e.g., using <code class="docutils literal"><span class="pre">{&quot;image_a&quot;:0,</span> <span class="pre">&quot;image_b&quot;:0,</span> <span class="pre">&quot;label&quot;:1}</span></code>) or skip item (e.g., using <code class="docutils literal"><span class="pre">{&quot;image_a&quot;:0,</span> <span class="pre">&quot;label&quot;:2}</span></code>).</p> <p>Using a dictionary (<code class="docutils literal"><span class="pre">{&quot;image&quot;:0,</span> <span class="pre">&quot;label&quot;:1}</span></code>) instead of a list (<code class="docutils literal"><span class="pre">[&quot;image&quot;,</span> <span class="pre">&quot;label&quot;]</span></code>) gives the advantage that the user can easily reuse the items (e.g., using <code class="docutils literal"><span class="pre">{&quot;image_a&quot;:0,</span> <span class="pre">&quot;image_b&quot;:0,</span> <span class="pre">&quot;label&quot;:1}</span></code>) or even skip an item (e.g., using <code class="docutils literal"><span class="pre">{&quot;image_a&quot;:0,</span> <span class="pre">&quot;label&quot;:2}</span></code>).</p>
</div> </div>
<div class="section" id="how-to-create-custom-data-reader-creator"> <div class="section" id="how-to-create-a-custom-data-reader-creator">
<span id="how-to-create-custom-data-reader-creator"></span><h3>How to create custom data reader creator<a class="headerlink" href="#how-to-create-custom-data-reader-creator" title="永久链接至标题"></a></h3> <span id="how-to-create-a-custom-data-reader-creator"></span><h3>How to create a custom data reader creator ?<a class="headerlink" href="#how-to-create-a-custom-data-reader-creator" title="永久链接至标题"></a></h3>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">image_reader_creator</span><span class="p">(</span><span class="n">image_path</span><span class="p">,</span> <span class="n">label_path</span><span class="p">,</span> <span class="n">n</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">image_reader_creator</span><span class="p">(</span><span class="n">image_path</span><span class="p">,</span> <span class="n">label_path</span><span class="p">,</span> <span class="n">n</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">reader</span><span class="p">():</span> <span class="k">def</span> <span class="nf">reader</span><span class="p">():</span>
<span class="n">f</span> <span class="o">=</span> <span class="nb">open</span><span class="p">(</span><span class="n">image_path</span><span class="p">)</span> <span class="n">f</span> <span class="o">=</span> <span class="nb">open</span><span class="p">(</span><span class="n">image_path</span><span class="p">)</span>
...@@ -377,7 +378,7 @@ ...@@ -377,7 +378,7 @@
</div> </div>
<div class="section" id="how-is-paddle-train-implemented"> <div class="section" id="how-is-paddle-train-implemented">
<span id="how-is-paddle-train-implemented"></span><h3>How is <code class="docutils literal"><span class="pre">paddle.train</span></code> implemented<a class="headerlink" href="#how-is-paddle-train-implemented" title="永久链接至标题"></a></h3> <span id="how-is-paddle-train-implemented"></span><h3>How is <code class="docutils literal"><span class="pre">paddle.train</span></code> implemented<a class="headerlink" href="#how-is-paddle-train-implemented" title="永久链接至标题"></a></h3>
<p>An example implementation of paddle.train could be:</p> <p>An example implementation of paddle.train is:</p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">train</span><span class="p">(</span><span class="n">batch_reader</span><span class="p">,</span> <span class="n">mapping</span><span class="p">,</span> <span class="n">batch_size</span><span class="p">,</span> <span class="n">total_pass</span><span class="p">):</span> <div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">train</span><span class="p">(</span><span class="n">batch_reader</span><span class="p">,</span> <span class="n">mapping</span><span class="p">,</span> <span class="n">batch_size</span><span class="p">,</span> <span class="n">total_pass</span><span class="p">):</span>
<span class="k">for</span> <span class="n">pass_idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">total_pass</span><span class="p">):</span> <span class="k">for</span> <span class="n">pass_idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">total_pass</span><span class="p">):</span>
<span class="k">for</span> <span class="n">mini_batch</span> <span class="ow">in</span> <span class="n">batch_reader</span><span class="p">():</span> <span class="c1"># this loop will never end in online learning.</span> <span class="k">for</span> <span class="n">mini_batch</span> <span class="ow">in</span> <span class="n">batch_reader</span><span class="p">():</span> <span class="c1"># this loop will never end in online learning.</span>
......
develop/doc_cn/searchindex.js
浏览文件 @ 02a44c38
因为 它太大了无法显示 source diff 。你可以改为 查看blob。
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