Luo Tao
8 years ago
commit
368737e41b
@ -0,0 +1,49 @@
|
||||
include(ExternalProject)
|
||||
|
||||
set(NCCL_SOURCE_DIR ${THIRD_PARTY_PATH}/nccl)
|
||||
|
||||
include_directories(${NCCL_SOURCE_DIR}/src/extern_nccl/src)
|
||||
|
||||
if(WITH_DSO)
|
||||
# If we use DSO, we do not build nccl, just download the dependencies
|
||||
set(NCCL_BUILD_COMMAND "")
|
||||
set(NCCL_INSTALL_COMMAND "")
|
||||
set(NCCL_INSTALL_DIR "")
|
||||
else()
|
||||
# otherwise, we build nccl and link it.
|
||||
set(NCCL_INSTALL_DIR ${THIRD_PARTY_PATH}/install/nccl)
|
||||
# Note: cuda 8.0 is needed to make nccl
|
||||
# When cuda is not installed on the system directory, need to set CUDA_HOME to your cuda root
|
||||
set(NCCL_BUILD_COMMAND "make -j 8")
|
||||
set(NCCL_INSTALL_COMMAND "make install PREFIX=${NCCL_INSTALL_DIR}")
|
||||
endif()
|
||||
|
||||
ExternalProject_Add(
|
||||
extern_nccl
|
||||
${EXTERNAL_PROJECT_LOG_ARGS}
|
||||
GIT_REPOSITORY "https://github.com/NVIDIA/nccl.git"
|
||||
GIT_TAG "v1.3.4-1"
|
||||
PREFIX "${NCCL_SOURCE_DIR}"
|
||||
UPDATE_COMMAND ""
|
||||
CONFIGURE_COMMAND ""
|
||||
BUILD_COMMAND "${NCCL_BUILD_COMMAND}"
|
||||
INSTALL_COMMAND "${NCCL_INSTALL_COMMAND}"
|
||||
INSTALL_DIR "${NCCL_INSTALL_DIR}"
|
||||
TEST_COMMAND ""
|
||||
)
|
||||
|
||||
if(WITH_DSO)
|
||||
if(${CMAKE_VERSION} VERSION_LESS "3.3.0")
|
||||
set(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/lib_nccl_dummy.c)
|
||||
file(WRITE ${dummyfile} "const char * dummy_nccl = \"${dummyfile}\";")
|
||||
add_library(nccl STATIC ${dummyfile})
|
||||
else()
|
||||
add_library(nccl INTERFACE)
|
||||
endif()
|
||||
else()
|
||||
add_library(nccl STATIC IMPORTED GLOBAL)
|
||||
set_property(TARGET nccl PROPERTY IMPORTED_LOCATION
|
||||
${NCCL_INSTALL_DIR}/lib/libnccl_static.a)
|
||||
endif()
|
||||
|
||||
add_dependencies(nccl extern_nccl)
|
||||
Binary file not shown.
@ -0,0 +1,232 @@
|
||||
## Survey on Graph
|
||||
|
||||
Neural network framework often provides symbolic API for users to write network topology conveniently. This doc manily focus on symbolic API in most popular neural network frameworks, and try to find out how to parse symbolic configuration to a portable file, such as protobuf or json.
|
||||
|
||||
### Mxnet
|
||||
|
||||
The core concept of symbolic API is `Symbol`. Mxnet implements `Symbol` class in C++, and export to Python using C-API. Please refer to the comments in Mxnet:
|
||||
|
||||
|
||||
`Symbol` is help class used to represent the operator node in Graph.
|
||||
`Symbol` acts as an interface for building graphs from different components like Variable, Functor and Group. `Symbol` is also exported to python front-end (while Graph is not) to enable quick test and deployment. Conceptually, symbol is the final operation of a graph and thus including all the information required (the graph) to evaluate its output value.
|
||||
|
||||
|
||||
A simple network topology wrote by Symbol is as follows:
|
||||
|
||||
```python
|
||||
def get_symbol(num_classes=10, **kwargs):
|
||||
data = mx.symbol.Variable('data')
|
||||
data = mx.symbol.Flatten(data=data)
|
||||
fc1 = mx.symbol.FullyConnected(data = data, name='fc1', num_hidden=128)
|
||||
act1 = mx.symbol.Activation(data = fc1, name='relu1', act_type="relu")
|
||||
fc2 = mx.symbol.FullyConnected(data = act1, name = 'fc2', num_hidden = 64)
|
||||
act2 = mx.symbol.Activation(data = fc2, name='relu2', act_type="relu")
|
||||
fc3 = mx.symbol.FullyConnected(data = act2, name='fc3', num_hidden=num_classes)
|
||||
mlp = mx.symbol.SoftmaxOutput(data = fc3, name = 'softmax')
|
||||
return mlp
|
||||
```
|
||||
|
||||
|
||||
|
||||
Varible here is actually a Symbol. Every basic Symbol will correspond to one Node, and every Node has its own NodeAttr. There is a op field in NodeAttr class, when a Symbol represents Variable(often input data), the op field is null.
|
||||
|
||||
Symbol contains a data member, std::vector<NodeEntry> outputs, and NodeEntry cantains a poniter to Node. We can follow the Node pointer to get all the Graph.
|
||||
|
||||
And Symbol can be saved to a Json file.
|
||||
|
||||
Here is a detailed example:
|
||||
|
||||
```
|
||||
>>> import mxnet as mx
|
||||
>>> data = mx.symbol.Variable('data')
|
||||
>>> print data.debug_str()
|
||||
Variable:data
|
||||
|
||||
>>> data = mx.symbol.Flatten(data=data)
|
||||
>>> print data.debug_str()
|
||||
Symbol Outputs:
|
||||
output[0]=flatten0(0)
|
||||
Variable:data
|
||||
--------------------
|
||||
Op:Flatten, Name=flatten0
|
||||
Inputs:
|
||||
arg[0]=data(0) version=0
|
||||
|
||||
>>> fc1 = mx.symbol.FullyConnected(data = data, name='fc1', num_hidden=128)
|
||||
>>> print fc1.debug_str()
|
||||
Symbol Outputs:
|
||||
output[0]=fc1(0)
|
||||
Variable:data
|
||||
--------------------
|
||||
Op:Flatten, Name=flatten0
|
||||
Inputs:
|
||||
arg[0]=data(0) version=0
|
||||
Variable:fc1_weight
|
||||
Variable:fc1_bias
|
||||
--------------------
|
||||
Op:FullyConnected, Name=fc1
|
||||
Inputs:
|
||||
arg[0]=flatten0(0)
|
||||
arg[1]=fc1_weight(0) version=0
|
||||
arg[2]=fc1_bias(0) version=0
|
||||
Attrs:
|
||||
num_hidden=128
|
||||
|
||||
```
|
||||
|
||||
|
||||
### TensorFlow
|
||||
|
||||
|
||||
The core concept of symbolic API is `Tensor`. Tensorflow defines `Tensor` in Python. Please refer to the comments in TensorFlow:
|
||||
|
||||
A `Tensor` is a symbolic handle to one of the outputs of an `Operation`. It does not hold the values of that operation's output, but instead provides a means of computing those values in a TensorFlow [Session](https://www.tensorflow.org/api_docs/python/tf/Session).
|
||||
|
||||
A simple example is as follows:
|
||||
|
||||
```python
|
||||
# Build a dataflow graph.
|
||||
c = tf.constant([[1.0, 2.0], [3.0, 4.0]])
|
||||
d = tf.constant([[1.0, 1.0], [0.0, 1.0]])
|
||||
e = tf.matmul(c, d)
|
||||
|
||||
# Construct a `Session` to execute the graph.
|
||||
sess = tf.Session()
|
||||
|
||||
# Execute the graph and store the value that `e` represents in `result`.
|
||||
result = sess.run(e)
|
||||
```
|
||||
|
||||
|
||||
The main method of `Tensor` is as follows:
|
||||
|
||||
|
||||
```python
|
||||
@property
|
||||
def op(self):
|
||||
"""The `Operation` that produces this tensor as an output."""
|
||||
return self._op
|
||||
|
||||
@property
|
||||
def dtype(self):
|
||||
"""The `DType` of elements in this tensor."""
|
||||
return self._dtype
|
||||
|
||||
@property
|
||||
def graph(self):
|
||||
"""The `Graph` that contains this tensor."""
|
||||
return self._op.graph
|
||||
|
||||
@property
|
||||
def name(self):
|
||||
"""The string name of this tensor."""
|
||||
if not self._op.name:
|
||||
raise ValueError("Operation was not named: %s" % self._op)
|
||||
return "%s:%d" % (self._op.name, self._value_index)
|
||||
|
||||
@property
|
||||
def device(self):
|
||||
"""The name of the device on which this tensor will be produced, or None."""
|
||||
return self._op.device
|
||||
```
|
||||
|
||||
|
||||
Tensor can be taken as target to run by session. Tensor contains all the information of Graph, and tracks data dependency.
|
||||
|
||||
|
||||
Here is a detailed example:
|
||||
|
||||
|
||||
```
|
||||
>>> import tensorflow as tf
|
||||
>>> c = tf.constant([[1.0, 2.0], [3.0, 4.0]])
|
||||
>>> print c.graph
|
||||
<tensorflow.python.framework.ops.Graph object at 0x10f256d50>
|
||||
>>> d = tf.constant([[1.0, 1.0], [0.0, 1.0]])
|
||||
>>> print d.graph
|
||||
<tensorflow.python.framework.ops.Graph object at 0x10f256d50>
|
||||
>>> e = tf.matmul(c, d)
|
||||
>>> print e.graph
|
||||
<tensorflow.python.framework.ops.Graph object at 0x10f256d50>
|
||||
```
|
||||
|
||||
### Dynet
|
||||
|
||||
|
||||
The core concept of symbolic API is `Expression`, and Dynet defines `Expression` class in C++.
|
||||
|
||||
|
||||
A simple example is as follows:
|
||||
|
||||
```cpp
|
||||
ComputationGraph cg;
|
||||
Expression W = parameter(cg, pW);
|
||||
|
||||
Expression in = input(cg, xs[i]);
|
||||
Expression label = input(cg, ys[i]);
|
||||
Expression pred = W * in;
|
||||
Expression loss = square(pred - label);
|
||||
```
|
||||
|
||||
The input data and parameter are also represented by Expression. Every basci Expression corresponds to a Node. And input data is also a Node.
|
||||
|
||||
Expression has a data member ComputationGraph, and ComputationGraph will be modified in users' configuring process. Expression can be a running target, beacuse Expression contains all dependency.
|
||||
|
||||
|
||||
Here is a detailed example:
|
||||
|
||||
write topology in C++
|
||||
|
||||
```
|
||||
ComputationGraph cg;
|
||||
Expression W = parameter(cg, pW);
|
||||
cg.print_graphviz();
|
||||
|
||||
Expression pred = W * xs[i];
|
||||
cg.print_graphviz();
|
||||
|
||||
Expression loss = square(pred - ys[i]);
|
||||
cg.print_graphviz();
|
||||
```
|
||||
|
||||
compile and print
|
||||
|
||||
```
|
||||
# first print
|
||||
digraph G {
|
||||
rankdir=LR;
|
||||
nodesep=.05;
|
||||
N0 [label="v0 = parameters({1}) @ 0x7ffe4de00110"];
|
||||
}
|
||||
# second print
|
||||
digraph G {
|
||||
rankdir=LR;
|
||||
nodesep=.05;
|
||||
N0 [label="v0 = parameters({1}) @ 0x7ffe4de00110"];
|
||||
N1 [label="v1 = v0 * -0.98"];
|
||||
N0 -> N1;
|
||||
}
|
||||
# third print
|
||||
digraph G {
|
||||
rankdir=LR;
|
||||
nodesep=.05;
|
||||
N0 [label="v0 = parameters({1}) @ 0x7ffe4de00110"];
|
||||
N1 [label="v1 = v0 * -0.98"];
|
||||
N0 -> N1;
|
||||
N2 [label="v2 = -1.88387 - v1"];
|
||||
N1 -> N2;
|
||||
N3 [label="v3 = -v2"];
|
||||
N2 -> N3;
|
||||
N4 [label="v4 = square(v3)"];
|
||||
N3 -> N4;
|
||||
}
|
||||
```
|
||||
|
||||
### Conclusion
|
||||
|
||||
|
||||
Actually, Symbol/Tensor/Expression in Mxnet/TensorFlow/Dynet are the same level concepts. We use a unified name Expression here, this level concept has following features:
|
||||
|
||||
- Users wirte topoloy with symbolic API, and all return value is Expression, including input data and parameter.
|
||||
- Expression corresponds with a global Graph, and Expression can also be composed.
|
||||
- Expression tracks all dependency and can be taken as a run target
|
||||
|
After Width: | Height: | Size: 32 KiB |
|
After Width: | Height: | Size: 45 KiB |
|
After Width: | Height: | Size: 1.1 KiB |
|
After Width: | Height: | Size: 989 B |
|
After Width: | Height: | Size: 1.6 KiB |
@ -0,0 +1,38 @@
|
||||
# Design Doc: Model Format
|
||||
|
||||
## Motivation
|
||||
|
||||
A model is an output of the training process. One complete model consists of two parts, the **topology** and the **parameters**. In order to support industrial deployment, the model format must be self-complete and must not expose any training source code.
|
||||
|
||||
As a result, In PaddlePaddle, the **topology** is represented as a [ProgramDesc](https://github.com/PaddlePaddle/Paddle/blob/1c0a4c901c9fc881d120249c703b15d1c50dae7d/doc/design/program.md), which describes the model structure. The **parameters** contain all the trainable weights in the model. We must support large size parameters and efficient serialization/deserialization of parameters.
|
||||
|
||||
## Implementation
|
||||
|
||||
The topology is saved as a plain text in a detailed self-contain protobuf file.
|
||||
|
||||
The parameters are saved as a binary file. As we all know, the protobuf message has a limit of [64M size](https://developers.google.com/protocol-buffers/docs/reference/cpp/google.protobuf.io.coded_stream#CodedInputStream.SetTotalBytesLimit.details). We have done a [benchmark experiment](https://github.com/PaddlePaddle/Paddle/pull/4610), which shows that protobuf is not fit for the task.
|
||||
|
||||
As a result, we design a particular format for tensor serialization. By default, an arbitrary tensor in Paddle is a [LoDTensor](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/lod_tensor.md), and has a description information proto of [LoDTensorDesc](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/framework.proto#L99). We save the DescProto as the byte string header. It contains all the necessary information, such as the `dims`, the `name` of the tensor, and the `LoD` information in [LoDTensor](https://github.com/PaddlePaddle/Paddle/blob/1c0a4c901c9fc881d120249c703b15d1c50dae7d/paddle/framework/lod_tensor.md). A tensor stores values in a continuous memory buffer. For speed we dump the raw memory to disk and save it as the byte string content. So, the binary format of one tensor is,
|
||||
|
||||
|HeaderLength|ContentLength|**LoDTensorDesc**|**TensorValue**|
|
||||
|
||||
The table below shows a tensor's byte view in detail. Note that all the signed values are written in the little-endian format.
|
||||
|
||||
```text
|
||||
[offset] [type] [description]
|
||||
0004 4 bytes integer HeaderLength, the length of LoDTensorDesc
|
||||
0008 4 bytes integer ContentLength, the length of LodTensor Buffer
|
||||
0009 1 bytes char TensorDesc
|
||||
00010 1 bytes char TensorDesc
|
||||
...
|
||||
00100 1 bytes char TensorValue
|
||||
00101 1 bytes char TensorValue
|
||||
00102 1 bytes char TensorValue ..
|
||||
...
|
||||
```
|
||||
|
||||
## Summary
|
||||
|
||||
- We introduce a model format.
|
||||
- The model represented by its forward-pass computation procedure is saved in a **ProgramDesc** protobuf message.
|
||||
- A bunch of specified format binary tensors describe the **parameters**.
|
||||
@ -0,0 +1,63 @@
|
||||
# Prune
|
||||
|
||||
## Motivation
|
||||
|
||||
We want to support running inference, training and checkpointing in one `ProgramDesc`. We implement
|
||||
`void Prune(const ProgramDesc* input, ProgramDesc* output)` function, which takes a `ProgramDesc`
|
||||
and generate a pruned `ProgramDesc`.
|
||||
|
||||
## Challenge
|
||||
|
||||
Pruning need to support both variables and operators being evaluation targets. Consider the following
|
||||
different situations.
|
||||
|
||||
```python
|
||||
# Case 1: run foward pass.
|
||||
cost_np = session.run(target=cost)
|
||||
# Case 2: run backward passing.
|
||||
opts_np, _ = session.run(target=[cost, opt])
|
||||
# Case 3: run checkpointing
|
||||
_ = session.run(target=checkpoint)
|
||||
```
|
||||
|
||||
## Solution
|
||||
|
||||
To support evaluation of operators, we add `is_target` field in the `OpDesc`.
|
||||
|
||||
```c++
|
||||
message OpDesc {
|
||||
required string type = 3;
|
||||
repeated Var inputs = 1;
|
||||
repeated Var outputs = 2;
|
||||
repeated Attr attrs = 4;
|
||||
optional bool is_target = 5 [ default = false ];
|
||||
};
|
||||
```
|
||||
|
||||
To support evaluation of variables, we add [fetch_op](https://github.com/PaddlePaddle/Paddle/pull/4599).
|
||||
For each variable in the `target`, we insert a `fetch_op` into the `ProgramDesc` with `variable` being
|
||||
`fetch_op`'s input. Then we also set `fetch_op` is a target.
|
||||
|
||||
### Algorithm
|
||||
|
||||
If an operator needs to be run, it must fall into one of the following cases:
|
||||
|
||||
1. It is the target.
|
||||
2. It is depended by some other ops, meaning its output is some other op's input.
|
||||
|
||||
The first case can be checked by `op_desc.is_traget()` . The second case can be implement as
|
||||
|
||||
```c++
|
||||
bool HasDependentVar(const OpDesc& op_desc, const std::set<string>& dependent_vars) {
|
||||
for (auto& var : op_desc.outputs()) {
|
||||
for (auto& argu : var.arguments()) {
|
||||
if (dependent_vars.count(argu) != 0) {
|
||||
return true;
|
||||
}
|
||||
}
|
||||
}
|
||||
return false;
|
||||
}
|
||||
```
|
||||
|
||||
Then the whole algorithm can be implemented as the following [code](https://github.com/tonyyang-svail/Paddle/blob/prune_impl/paddle/framework/prune.cc).
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
Some files were not shown because too many files have changed in this diff Show More
Loading…
Reference in new issue