Skip to content

P
Paddle

PaddlePaddle / Paddle
1 年多 前同步成功

通知 2302
Star 20931
Fork 5422
P
Paddle
提交 5a1f0617 编写于 作者: Y Yi Wang
浏览文件
操作

Update according to discussions in https://github.com/PaddlePaddle/Paddle/issues/1315

上级 a30f6aa9
显示空白变更内容
内联 并排
Showing with 119 addition and 174 deletion
doc/design/api.md
浏览文件 @ 5a1f0617
...@@ -2,140 +2,148 @@ ...@@ -2,140 +2,148 @@
## Ingredients ## Ingredients
As the first step of our design, we list important concepts in deep As our design principle is starting from the essence: how could we
learning and try to figure their relationship, as shown below: allow users to express and solve their problems at neural networks.
Some essential concepts that our API have to provide include:
``` 1. A *topology* is an expression of *layers*.
Model = {topology, parameters}
Evaluator = {Model*, activations}
- forward
- test(cost, ...)
GradientMachine = {Evaluator*, gradients} 1. A layer could be any kind of computation, including *cost*.
- backward
Optimizer = {GradientMachine*} 1. Some layers have parameters, some don't. Most costs don't have
- train(cost, ...) parameters.
- update
- checkpoint
```
where the pair of curly braces `{` and `}` indicate *composition*, `*` 1. In some topologies, layers share parameters. For
indicates a *reference*, and `-` marks a "class method". example,
[the network for training a ranking model](https://github.com/PaddlePaddle/Paddle/issues/1311#issuecomment-279121850).
1. At programming time, users specify topologies and possible sharing
of parameters. PaddlePaddle can figure out and create parameters
required (and possibly shared) by one or more topologies.
### Model
We used to think that parameters are part of the topology (or layers). ## Starting from Examples
But that is not true because multiple layers could share the same
parameter matrix. An example is a network that compares two text
segments in a semantic space:
``` As a summarization
semantic of
text A -> projection ---\ [our disucssion](https://github.com/PaddlePaddle/Paddle/issues/1315),
layer A \ let us present two examples here:
cosine
similarity -> output
layer
semantic /
text B -> projection ---/
layer B
```
In this network, the two semantic projection layers (A and B) share
the same parameter matrix.
For more information about our API that specifies topology and ### Example 1. Sharing Parameters between Layers
parameter sharing, please refer to [TODO: API].
We use
the
[3-branch ranking](https://github.com/PaddlePaddle/Paddle/issues/1311#issuecomment-279121850) model
in this example. For your convenience, I copy-a-paste the model's
topology as follows:
### Evaluator ```
A -> f -\
Q -> f --> cost
B -> f -/
```
Supposed that we have a trained ranking model, we should be able to The following program trains the topology including the cost, and then
use it in our search engine. The search engine's Web server is a use the sub-network in the trained topology in inference:
concurrent program so to serve many HTTP requests simultaneously. It
doesn't make sense for each of these threads to have its own copy of the model because that would duplicate topologies and parameters.
However, each thread should be able to record layer outputs, i.e.,
activations, computed from an input, derived from the request. With
*Evaluator* that saves activations, we can write the over-simplified
server program as:
```python ```python
m = paddle.model.load("trained.model") def f(in):
e = paddle.layer.embedding(in, parameter_name="embedding")
http.handle("/", o = paddle.layer.softmax(e, parameter_name="semantic")
lambda req: return o
e = paddle.evaluator.create(m)
e.forward(req) # Create 3 topologies (subnets), they share parameters because all
e.activation(layer="output")) # returns activations of layer "output" # correspoinding layers have the same parameter names.
fA = f(paddle.layer.data(input_name="A"))
fB = f(paddle.layer.data(input_name="B"))
fQ = f(paddle.layer.data(input_name="Q"))
topology = paddle.cost.less_than(
paddle.cost.cross_entropy(fA, fQ),
paddle.cost.corss_entropy(fB, fQ))
# Derive parameters required in topology and create them in model.
parameters = paddle.parameters.create(topology)
# Estimate parameters used in topology from data.
paddle.train(topology, parameters, reader=read_ranking_model_data)
# Inference using fA (or fB or fC, as they share their parameters).
[testA, testB, testQ] = read_ranking_model_data()
print "The sematic-vector of testA: ", paddle.infer(fA, parameters, testA)
``` ```
### GradientMachine
Similar to the evaluation, the training needs to compute gradients so
to update model parameters. Because an [optimizer](#optimizer) might
run multiple simultaneous threads to update the same model, gradients
should be separated from the model. Because gradients are only used
in training, but not serving, they should be separate from Evaluator.
Hence the `GradientMachine`.
### Optimizer ### Example 2. Sharing Parameters between "Models"
None of Model, Evaluator, nor GradientMachine implements the training We use [GAN](https://github.com/PaddlePaddle/book/tree/develop/gan) in
loop, hence Optimizer. We can define a concurrent optimizer that runs this example. In the following example program, `d0` and `d1`
multiple simultaneous threads to train a model -- just let each correspond to the two networks in the following figure:
thread has its own GradientMachine object.
Most models should be able to be trained using the <img src="https://github.com/wangyang59/book/raw/00036f4b0da5225041a6824587c1a01cf20159b1/gan/image/gan_ig.png" width=400 />
`paddle.optimizer.SGD` by calling its `train` method. Many
customizations to the SGD algorithm happens with the update equation,
e.g., momentum and the Adam SGD algorithm. We make `train` calls
`update` to do an update, so that we can derive a `paddle.optimizer.Adam`
from `paddle.optimizer.SGD` by overrides only the `update` method.
## Programming Interface
A fictive example of PaddlePaddle program looks like the following:
```python ```python
import paddle def G(in):
# over-simplified example as G has only one layers:
return paddle.layer.fc(in, parameter_name="G")
def D(in);
# again, over-simplified:
return paddle.layer.fc(in, parameter_name="D")
# Construct the first topology, which contains both D and G.
# By learning this topology, we update parameters of G.
d0 = paddle.cost.should_be_false(D(G(paddle.layer.data())))
# Construct a second topology d1, which contains only D. By
# training this topology, we update parameters of D. Note
# that d1 share parameters with d0.
d1 = paddle.cost.should_be_true(D(paddle.layer.data()))
# Create parameters from a list of multiple topologies (models) for
# the chance to share parameters between these topologies.
parameters = paddle.parameters.create([d0, d1])
# Iterative training of GAN.
for ...:
train(d0, parameters, reader=read_from_rng, immutable_parameters={"D"})
train(d1, parameters, reader=read_from_realistic_images)
# Use d1 for inference:
print "D thinks a batch of images are realistic ", infer(d1, parameters, read_mnist_images)
```
def read(args):
f = open_file(args["filename"])
mb = read_a_minibatch(f)
end_pass = eof(f)
if end_pass:
f = open_file(args["filename"]) # rewind for reading again
yield mb, end_pass
input = paddle.layer.data(...) ### Summarization
intermediate = paddle.layers.fc(input)
output = paddle.layer.softmax(intermediate)
model = paddle.model.create(output)
paddle.train(model, data_provider=read) Above two programs reveal some important design concerns:
```
This shows some important part of a program: 1. Users describe a topology as an expression of layers. Every layer
has a *parameter name*. If the users don't specify it explicitly, it's automatically generated as a unique name. By
specifying the parameter name, users can specify the sharing of
parameters between layers and even between topologies.
1. Define how to read (and augment) data by defining a function, in 1. `paddle.parameters.create` figures out parameters required by one
this example, `read`, that `yields` a minibatch and a boolean flag or more topologies from parameter names of layers. It creates these
`eof_of_pass`. parameters and returns a `ParameterSet` object, which is in essence
a map from *parameter names* to *parameters*.
1. Define the topology, `input`, `intermediate`, and `output` in this 1. At training and inference time, `paddle.train` and `paddle.infer`
example. requires both a topology and the parameter set that holds the parameters of that topology. There are some reasons:
1. Create parameters from the topology thus forms the model by calling 1. This prevents users from forgetting to call
`paddel.model.create`. `paddle.parameters.create`.
1. `paddle.train` needs to know which parameter set to update.
1. Users could load another (pre-trained) parameter set and use it
with a topology in `train.infer`.
1. Train the model by calling `paddle.train`. 1. By specifying the `immutable_parameters` parameter of
`paddle.train`, we can forbid the update of these parameters.
### Reader ## Reader
Not all programming frameworks allow users to define I/O functions. Not all programming frameworks allow users to define I/O functions.
An example is Google MapReduce, which can only read from text, An example is Google MapReduce, which can only read from text,
...@@ -145,78 +153,15 @@ readers and writers by deriving from base classes `Reader` and ...@@ -145,78 +153,15 @@ readers and writers by deriving from base classes `Reader` and
decide to provide the flexibility to users to define their readers. decide to provide the flexibility to users to define their readers.
#### A Synthetic Data Reader There are some open questions here:
Sometimes we want to test a topology and/or a training algorithm using
synthetic data. We can do this by defining the reader a synthesizer:
```python 1. **Should a reader return a Python dictionary?**
def read(args):
x = sample_from_uniform(0.0, 1.0)
y = sample_from_gauss(2 * x, sigma)
yield {x, y}, False # no end-of-file so no end-of-pass
```
#### A Reader for Online Learning 1. **How to map multiple outputs from a reader to multiple data layers?**
Readers can also read an infinite data stream, e.g., a log stream from 1. **How to easily compose some existing readers to read more data and
a search engine and collected by Kafka: feed a topology with more data layers?**
```python
def read(args):
log_stream = kafka.open_channel(args["kafka channel name"])
yeild log_stream.read(), False # no end-of-pass in online learning
```
### Topology
By default, layers don't have names. But if we want to refer to a
layer later some time, for example, when we do serving using the model
and wants activations/outputs of a layer, we should give it a name.
```python
input = paddle.layer.data(...)
intermediate = paddle.layer.fc(input, name="inter", ...)
output = paddle.layer.softmax(intermediate, name="output", ...)
m = paddle.model.create(output)
e = paddle.evaluator.create(model)
e.forward(read_an_input()) # compute activations of all layers.
print e.activations(layer="inter") # retrieve the activations of layer "inter"
print e.activations(layer="output") # retrieve the activations of layer "output"
```
#### Sharing Parameters
In [above section](#model) we shows a network whose two layers share
the same parameter matrix. To specify such cases, we give "parameter
names" to layers. If some layers have the same paraemter names,
`paddle.model.create` creates a single parameter matrix for these
layers:
```python
text1 = paddle.layer.data(...)
sematic1 = paddle.layer.fc(text1, ..., parameter_name="sematic_projection")
text2 = paddle.layer.data(...)
sematic2 = paddle.layer.fc(text2, ..., parameter_name="sematic_projection")
out = paddle.layer.cosine(semantic1, semantic2)
```
We can also share parameter matrices between layers in different
models. To do this, we need an additional parameter that refers to a
model:
```python
model1_input = paddle.layer.data(...)
model1_output = paddle.layer.softmax(model1_input, ...,
parameter_name="a_parameter_matrix")
model1 = paddle.model.create(model1_output)
# Another model
model2_semantic = paddle.layer.fc(text2, ...,
parameter_name="a_parameter_matrix",
parameter_model=model1)
```
### Training ### Training
...@@ -270,7 +215,7 @@ def dist_train(): ...@@ -270,7 +215,7 @@ def dist_train():
Please be aware that if a process is running on the Kubernetes Please be aware that if a process is running on the Kubernetes
cluster, it will have some environment variables pre-defined. cluster, it will have some environment variables pre-defined.
If `dist_train` doesn't see these environment variables, it knowns If `dist_train` doesn't see these environment variables, it knows
that it's running on users' personal computer, and it should work as a that it's running on users' personal computer, and it should work as a
*launcher*. Otherwise, it knows that it's running on the cluster and *launcher*. Otherwise, it knows that it's running on the cluster and
need to figure out its role as either the master, or a trainer, or a need to figure out its role as either the master, or a trainer, or a
......
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