Skip to content

P
PaddleDetection

PaddlePaddle / PaddleDetection
接近 2 年 前同步成功

通知 707
Star 11112
Fork 2696
P
PaddleDetection
提交 e4880016 编写于 作者: L liaogang
浏览文件
操作

Merge conflict with hl_cuda_device.cc

上级 8393c19c 3a982d33
展开全部 隐藏空白更改
内联 并排
Showing with 12144 addition and 8902 deletion
.pre-commit-config.yaml
浏览文件 @ e4880016
......@@ -7,18 +7,14 @@
hooks:
- id: yapf
- repo: https://github.com/pre-commit/pre-commit-hooks
sha: 4ef03c4223ad322c7adaa6c6c0efb26b57df3b71
sha: 7539d8bd1a00a3c1bfd34cdb606d3a6372e83469
hooks:
- id: check-added-large-files
- id: check-merge-conflict
- id: check-symlinks
- id: detect-private-key
- id: end-of-file-fixer
# TODO(yuyang): trailing whitespace has some bugs on markdown
# files now, please not add it to pre-commit hook now
# - id: trailing-whitespace
#
# TODO(yuyang): debug-statements not fit for Paddle, because
# not all of our python code is runnable. Some are used for
# documenation
# - id: debug-statements
- repo: https://github.com/PaddlePaddle/clang-format-pre-commit-hook.git
sha: 28c0ea8a67a3e2dbbf4822ef44e85b63a0080a29
hooks:
- id: clang-formater
CMakeLists.txt
浏览文件 @ e4880016
......@@ -2,8 +2,8 @@ cmake_minimum_required(VERSION 2.8)
project(paddle CXX C)
set(PADDLE_MAJOR_VERSION 0)
set(PADDLE_MINOR_VERSION 8)
set(PADDLE_PATCH_VERSION 0b3)
set(PADDLE_MINOR_VERSION 9)
set(PADDLE_PATCH_VERSION 0a0)
set(PADDLE_VERSION ${PADDLE_MAJOR_VERSION}.${PADDLE_MINOR_VERSION}.${PADDLE_PATCH_VERSION})
set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/cmake")
......
README.md
浏览文件 @ e4880016
# PaddlePaddle
[![Build Status](https://travis-ci.org/baidu/Paddle.svg?branch=master)](https://travis-ci.org/baidu/Paddle)
[![Coverage Status](https://coveralls.io/repos/github/baidu/Paddle/badge.svg?branch=develop)](https://coveralls.io/github/baidu/Paddle?branch=develop)
[![Join the chat at https://gitter.im/PaddlePaddle/Deep_Learning](https://badges.gitter.im/Join%20Chat.svg)](https://gitter.im/PaddlePaddle/Deep_Learning?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge)
[![License](https://img.shields.io/badge/license-Apache%202.0-green.svg)](LICENSE)
[![Build Status](https://travis-ci.org/PaddlePaddle/Paddle.svg?branch=develop)](https://travis-ci.org/PaddlePaddle/Paddle)
[![Documentation Status](https://img.shields.io/badge/docs-latest-brightgreen.svg?style=flat)](http://www.paddlepaddle.org/)
[![Documentation Status](https://img.shields.io/badge/中文文档-最新-brightgreen.svg)](http://www.paddlepaddle.org/cn/index.html)
[![Coverage Status](https://coveralls.io/repos/github/PaddlePaddle/Paddle/badge.svg?branch=develop)](https://coveralls.io/github/PaddlePaddle/Paddle?branch=develop)
[![Release](https://img.shields.io/github/release/PaddlePaddle/Paddle.svg)](https://github.com/PaddlePaddle/Paddle/releases)
[![License](https://img.shields.io/badge/license-Apache%202-blue.svg)](LICENSE)
Welcome to the PaddlePaddle GitHub.
......@@ -14,7 +17,7 @@ developed by Baidu scientists and engineers for the purpose of applying deep
learning to many products at Baidu.
Our vision is to enable deep learning for everyone via PaddlePaddle.
Please refer to our [release announcement](https://github.com/baidu/Paddle/releases) to track the latest feature of PaddlePaddle.
Please refer to our [release announcement](https://github.com/PaddlePaddle/Paddle/releases) to track the latest feature of PaddlePaddle.
## Features
......@@ -26,15 +29,15 @@ Please refer to our [release announcement](https://github.com/baidu/Paddle/relea
connection.
- **Efficiency**
In order to unleash the power of heterogeneous computing resource,
optimization occurs at different levels of PaddlePaddle, including
computing, memory, architecture and communication. The following are some
examples:
- Optimized math operations through SSE/AVX intrinsics, BLAS libraries
(e.g. MKL, ATLAS, cuBLAS) or customized CPU/GPU kernels.
- Highly optimized recurrent networks which can handle **variable-length**
(e.g. MKL, ATLAS, cuBLAS) or customized CPU/GPU kernels.
- Highly optimized recurrent networks which can handle **variable-length**
sequence without padding.
- Optimized local and distributed training for models with high dimensional
sparse data.
......@@ -57,41 +60,39 @@ Please refer to our [release announcement](https://github.com/baidu/Paddle/relea
## Installation
Check out the [Install Guide](http://paddlepaddle.org/doc/build/) to install from
pre-built packages (**docker image**, **deb package**) or
pre-built packages (**docker image**, **deb package**) or
directly build on **Linux** and **Mac OS X** from the source code.
## Documentation
Both [English Docs](http://paddlepaddle.org/doc/) and [Chinese Docs](http://paddlepaddle.org/doc_cn/) are provided for our users and developers.
- [Quick Start](http://paddlepaddle.org/doc/demo/quick_start/index_en) <br>
You can follow the quick start tutorial to learn how use PaddlePaddle
step-by-step.
- [Example and Demo](http://paddlepaddle.org/doc/demo/) <br>
We provide five demos, including: image classification, sentiment analysis,
sequence to sequence model, recommendation, semantic role labeling.
sequence to sequence model, recommendation, semantic role labeling.
- [Distributed Training](http://paddlepaddle.org/doc/cluster) <br>
This system supports training deep learning models on multiple machines
with data parallelism.
- [Python API](http://paddlepaddle.org/doc/ui/) <br>
PaddlePaddle supports using either Python interface or C++ to build your
system. We also use SWIG to wrap C++ source code to create a user friendly
interface for Python. You can also use SWIG to create interface for your
favorite programming language.
- [How to Contribute](http://paddlepaddle.org/doc/build/contribute_to_paddle.html) <br>
We sincerely appreciate your interest and contributions. If you would like to
contribute, please read the contribution guide.
contribute, please read the contribution guide.
- [Source Code Documents](http://paddlepaddle.org/doc/source/) <br>
## Ask Questions
Please join the [**gitter chat**](https://gitter.im/PaddlePaddle/Deep_Learning) or send email to
**paddle-dev@baidu.com** to ask questions and talk about methods and models.
Framework development discussions and
bug reports are collected on [Issues](https://github.com/baidu/paddle/issues).
You are welcome to submit questions and bug reports as [Github Issues](https://github.com/PaddlePaddle/Paddle/issues).
## Copyright and License
PaddlePaddle is provided under the [Apache-2.0 license](LICENSE).
demo/semantic_role_labeling/data/extract_dict_feature.py
浏览文件 @ e4880016
......@@ -17,24 +17,15 @@ import os
from optparse import OptionParser
def extract_dict_features(pair_file, feature_file, src_dict_file,
tgt_dict_file):
src_dict = set()
tgt_dict = set()
with open(pair_file) as fin, open(feature_file, 'w') as feature_out, open(
src_dict_file, 'w') as src_dict_out, open(tgt_dict_file,
'w') as tgt_dict_out:
def extract_dict_features(pair_file, feature_file):
with open(pair_file) as fin, open(feature_file, 'w') as feature_out:
for line in fin:
sentence, labels = line.strip().split('\t')
sentence, predicate, labels = line.strip().split('\t')
sentence_list = sentence.split()
labels_list = labels.split()
src_dict.update(sentence_list)
tgt_dict.update(labels_list)
verb_index = labels_list.index('B-V')
verb_feature = sentence_list[verb_index]
mark = [0] * len(labels_list)
if verb_index > 0:
......@@ -42,47 +33,50 @@ def extract_dict_features(pair_file, feature_file, src_dict_file,
ctx_n1 = sentence_list[verb_index - 1]
else:
ctx_n1 = 'bos'
ctx_n1_feature = ctx_n1
if verb_index > 1:
mark[verb_index - 2] = 1
ctx_n2 = sentence_list[verb_index - 2]
else:
ctx_n2 = 'bos'
mark[verb_index] = 1
ctx_0_feature = sentence_list[verb_index]
ctx_0 = sentence_list[verb_index]
if verb_index < len(labels_list) - 2:
mark[verb_index + 1] = 1
ctx_p1 = sentence_list[verb_index + 1]
else:
ctx_p1 = 'eos'
ctx_p1_feature = ctx_p1
if verb_index < len(labels_list) - 3:
mark[verb_index + 2] = 1
ctx_p2 = sentence_list[verb_index + 2]
else:
ctx_p2 = 'eos'
feature_str = sentence + '\t' \
+ verb_feature + '\t' \
+ ctx_n1_feature + '\t' \
+ ctx_0_feature + '\t' \
+ ctx_p1_feature + '\t' \
+ predicate + '\t' \
+ ctx_n2 + '\t' \
+ ctx_n1 + '\t' \
+ ctx_0 + '\t' \
+ ctx_p1 + '\t' \
+ ctx_p2 + '\t' \
+ ' '.join([str(i) for i in mark]) + '\t' \
+ labels
feature_out.write(feature_str + '\n')
src_dict_out.write('<unk>\n')
src_dict_out.write('\n'.join(list(src_dict)))
tgt_dict_out.write('\n'.join(list(tgt_dict)))
if __name__ == '__main__':
usage = '-p pair_file -f feature_file -s source dictionary -t target dictionary '
usage = '-p pair_file -f feature_file'
parser = OptionParser(usage)
parser.add_option('-p', dest='pair_file', help='the pair file')
parser.add_option(
'-f', dest='feature_file', help='the file to store feature')
parser.add_option(
'-s', dest='src_dict', help='the file to store source dictionary')
parser.add_option(
'-t', dest='tgt_dict', help='the file to store target dictionary')
parser.add_option('-f', dest='feature_file', help='the feature file')
(options, args) = parser.parse_args()
extract_dict_features(options.pair_file, options.feature_file,
options.src_dict, options.tgt_dict)
extract_dict_features(options.pair_file, options.feature_file)
demo/semantic_role_labeling/data/extract_pairs.py
浏览文件 @ e4880016
......@@ -51,7 +51,7 @@ def read_sentences(words_file):
for line in fin:
line = line.strip()
if line == '':
sentences.append(s.lower())
sentences.append(s)
s = ''
else:
s += line + ' '
......@@ -64,6 +64,11 @@ def transform_labels(sentences, labels):
if len(labels[i]) == 1:
continue
else:
verb_list = []
for x in labels[i][0]:
if x !='-':
verb_list.append(x)
for j in xrange(1, len(labels[i])):
label_list = labels[i][j]
current_tag = 'O'
......@@ -88,8 +93,7 @@ def transform_labels(sentences, labels):
is_in_bracket = True
else:
print 'error:', ll
sen_lab_pair.append((sentences[i], label_seq))
sen_lab_pair.append((sentences[i], verb_list[j-1], label_seq))
return sen_lab_pair
......@@ -97,9 +101,9 @@ def write_file(sen_lab_pair, output_file):
with open(output_file, 'w') as fout:
for x in sen_lab_pair:
sentence = x[0]
label_seq = ' '.join(x[1])
assert len(sentence.split()) == len(x[1])
fout.write(sentence + '\t' + label_seq + '\n')
label_seq = ' '.join(x[2])
assert len(sentence.split()) == len(x[2])
fout.write(sentence + '\t' + x[1]+'\t' +label_seq + '\n')
if __name__ == '__main__':
......
demo/semantic_role_labeling/data/get_data.sh
浏览文件 @ e4880016
......@@ -14,6 +14,10 @@
# limitations under the License.
set -e
wget http://www.cs.upc.edu/~srlconll/conll05st-tests.tar.gz
wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/semantic_role_labeling/verbDict.txt --no-check-certificate
wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/semantic_role_labeling/targetDict.txt --no-check-certificate
wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/semantic_role_labeling/wordDict.txt --no-check-certificate
wget https://www.googledrive.com/host/0B7Q8d52jqeI9ejh6Q1RpMTFQT1k/semantic_role_labeling/emb --no-check-certificate
tar -xzvf conll05st-tests.tar.gz
rm conll05st-tests.tar.gz
cp ./conll05st-release/test.wsj/words/test.wsj.words.gz .
......@@ -22,4 +26,4 @@ gunzip test.wsj.words.gz
gunzip test.wsj.props.gz
python extract_pairs.py -w test.wsj.words -p test.wsj.props -o test.wsj.seq_pair
python extract_dict_feature.py -p test.wsj.seq_pair -f feature -s src.dict -t tgt.dict
python extract_dict_feature.py -p test.wsj.seq_pair -f feature
demo/semantic_role_labeling/dataprovider.py
浏览文件 @ e4880016
......@@ -17,11 +17,15 @@ from paddle.trainer.PyDataProvider2 import *
UNK_IDX = 0
def hook(settings, word_dict, label_dict, **kwargs):
def hook(settings, word_dict, label_dict, predicate_dict, **kwargs):
settings.word_dict = word_dict
settings.label_dict = label_dict
settings.predicate_dict = predicate_dict
#all inputs are integral and sequential type
settings.slots = [
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(predicate_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
......@@ -31,27 +35,33 @@ def hook(settings, word_dict, label_dict, **kwargs):
]
@provider(init_hook=hook)
def process(obj, file_name):
def get_batch_size(yeild_data):
return len(yeild_data[0])
@provider(init_hook=hook, should_shuffle=True, calc_batch_size=get_batch_size,
can_over_batch_size=False, cache=CacheType.CACHE_PASS_IN_MEM)
def process(settings, file_name):
with open(file_name, 'r') as fdata:
for line in fdata:
sentence, predicate, ctx_n1, ctx_0, ctx_p1, mark, label = \
sentence, predicate, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2, mark, label = \
line.strip().split('\t')
words = sentence.split()
sen_len = len(words)
word_slot = [obj.word_dict.get(w, UNK_IDX) for w in words]
word_slot = [settings.word_dict.get(w, UNK_IDX) for w in words]
predicate_slot = [obj.word_dict.get(predicate, UNK_IDX)] * sen_len
ctx_n1_slot = [obj.word_dict.get(ctx_n1, UNK_IDX)] * sen_len
ctx_0_slot = [obj.word_dict.get(ctx_0, UNK_IDX)] * sen_len
ctx_p1_slot = [obj.word_dict.get(ctx_p1, UNK_IDX)] * sen_len
predicate_slot = [settings.predicate_dict.get(predicate)] * sen_len
ctx_n2_slot = [settings.word_dict.get(ctx_n2, UNK_IDX)] * sen_len
ctx_n1_slot = [settings.word_dict.get(ctx_n1, UNK_IDX)] * sen_len
ctx_0_slot = [settings.word_dict.get(ctx_0, UNK_IDX)] * sen_len
ctx_p1_slot = [settings.word_dict.get(ctx_p1, UNK_IDX)] * sen_len
ctx_p2_slot = [settings.word_dict.get(ctx_p2, UNK_IDX)] * sen_len
marks = mark.split()
mark_slot = [int(w) for w in marks]
label_list = label.split()
label_slot = [obj.label_dict.get(w) for w in label_list]
yield word_slot, predicate_slot, ctx_n1_slot, \
ctx_0_slot, ctx_p1_slot, mark_slot, label_slot
label_slot = [settings.label_dict.get(w) for w in label_list]
yield word_slot, predicate_slot, ctx_n2_slot, ctx_n1_slot, \
ctx_0_slot, ctx_p1_slot, ctx_p2_slot, mark_slot, label_slot
demo/semantic_role_labeling/db_lstm.py
浏览文件 @ e4880016
......@@ -18,8 +18,9 @@ import sys
from paddle.trainer_config_helpers import *
#file paths
word_dict_file = './data/src.dict'
label_dict_file = './data/tgt.dict'
word_dict_file = './data/wordDict.txt'
label_dict_file = './data/targetDict.txt'
predicate_file= './data/verbDict.txt'
train_list_file = './data/train.list'
test_list_file = './data/test.list'
......@@ -30,8 +31,10 @@ if not is_predict:
#load dictionaries
word_dict = dict()
label_dict = dict()
predicate_dict = dict()
with open(word_dict_file, 'r') as f_word, \
open(label_dict_file, 'r') as f_label:
open(label_dict_file, 'r') as f_label, \
open(predicate_file, 'r') as f_pre:
for i, line in enumerate(f_word):
w = line.strip()
word_dict[w] = i
......@@ -40,6 +43,11 @@ if not is_predict:
w = line.strip()
label_dict[w] = i
for i, line in enumerate(f_pre):
w = line.strip()
predicate_dict[w] = i
if is_test:
train_list_file = None
......@@ -50,91 +58,157 @@ if not is_predict:
module='dataprovider',
obj='process',
args={'word_dict': word_dict,
'label_dict': label_dict})
'label_dict': label_dict,
'predicate_dict': predicate_dict })
word_dict_len = len(word_dict)
label_dict_len = len(label_dict)
pred_len = len(predicate_dict)
else:
word_dict_len = get_config_arg('dict_len', int)
label_dict_len = get_config_arg('label_len', int)
pred_len = get_config_arg('pred_len', int)
############################## Hyper-parameters ##################################
mark_dict_len = 2
word_dim = 32
mark_dim = 5
hidden_dim = 128
hidden_dim = 512
depth = 8
emb_lr = 1e-2
fc_lr = 1e-2
lstm_lr = 2e-2
########################### Optimizer #######################################
settings(
batch_size=150,
learning_method=AdamOptimizer(),
learning_rate=1e-3,
learning_method=MomentumOptimizer(momentum=0),
learning_rate=2e-2,
regularization=L2Regularization(8e-4),
gradient_clipping_threshold=25)
is_async=False,
model_average=ModelAverage(average_window=0.5,
max_average_window=10000),
)
#6 features
####################################### network ##############################
#8 features and 1 target
word = data_layer(name='word_data', size=word_dict_len)
predicate = data_layer(name='verb_data', size=word_dict_len)
predicate = data_layer(name='verb_data', size=pred_len)
ctx_n2 = data_layer(name='ctx_n2_data', size=word_dict_len)
ctx_n1 = data_layer(name='ctx_n1_data', size=word_dict_len)
ctx_0 = data_layer(name='ctx_0_data', size=word_dict_len)
ctx_p1 = data_layer(name='ctx_p1_data', size=word_dict_len)
ctx_p2 = data_layer(name='ctx_p2_data', size=word_dict_len)
mark = data_layer(name='mark_data', size=mark_dict_len)
if not is_predict:
target = data_layer(name='target', size=label_dict_len)
ptt = ParameterAttribute(name='src_emb', learning_rate=emb_lr)
layer_attr = ExtraLayerAttribute(drop_rate=0.5)
fc_para_attr = ParameterAttribute(learning_rate=fc_lr)
lstm_para_attr = ParameterAttribute(initial_std=0., learning_rate=lstm_lr)
para_attr = [fc_para_attr, lstm_para_attr]
word_embedding = embedding_layer(size=word_dim, input=word, param_attr=ptt)
predicate_embedding = embedding_layer(
size=word_dim, input=predicate, param_attr=ptt)
ctx_n1_embedding = embedding_layer(size=word_dim, input=ctx_n1, param_attr=ptt)
ctx_0_embedding = embedding_layer(size=word_dim, input=ctx_0, param_attr=ptt)
ctx_p1_embedding = embedding_layer(size=word_dim, input=ctx_p1, param_attr=ptt)
mark_embedding = embedding_layer(size=mark_dim, input=mark)
default_std=1/math.sqrt(hidden_dim)/3.0
emb_para = ParameterAttribute(name='emb', initial_std=0., learning_rate=0.)
std_0 = ParameterAttribute(initial_std=0.)
std_default = ParameterAttribute(initial_std=default_std)
predicate_embedding = embedding_layer(size=word_dim, input=predicate, param_attr=ParameterAttribute(name='vemb',initial_std=default_std))
mark_embedding = embedding_layer(name='word_ctx-in_embedding', size=mark_dim, input=mark, param_attr=std_0)
word_input=[word, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2]
emb_layers = [embedding_layer(size=word_dim, input=x, param_attr=emb_para) for x in word_input]
emb_layers.append(predicate_embedding)
emb_layers.append(mark_embedding)
hidden_0 = mixed_layer(
name='hidden0',
size=hidden_dim,
input=[
full_matrix_projection(input=word_embedding),
full_matrix_projection(input=predicate_embedding),
full_matrix_projection(input=ctx_n1_embedding),
full_matrix_projection(input=ctx_0_embedding),
full_matrix_projection(input=ctx_p1_embedding),
full_matrix_projection(input=mark_embedding),
])
bias_attr=std_default,
input=[ full_matrix_projection(input=emb, param_attr=std_default ) for emb in emb_layers ])
lstm_0 = lstmemory(input=hidden_0, layer_attr=layer_attr)
mix_hidden_lr = 1e-3
lstm_para_attr = ParameterAttribute(initial_std=0.0, learning_rate=1.0)
hidden_para_attr = ParameterAttribute(initial_std=default_std, learning_rate=mix_hidden_lr)
lstm_0 = lstmemory(name='lstm0',
input=hidden_0,
act=ReluActivation(),
gate_act=SigmoidActivation(),
state_act=SigmoidActivation(),
bias_attr=std_0,
param_attr=lstm_para_attr)
#stack L-LSTM and R-LSTM with direct edges
input_tmp = [hidden_0, lstm_0]
for i in range(1, depth):
fc = fc_layer(input=input_tmp, size=hidden_dim, param_attr=para_attr)
mix_hidden = mixed_layer(name='hidden'+str(i),
size=hidden_dim,
bias_attr=std_default,
input=[full_matrix_projection(input=input_tmp[0], param_attr=hidden_para_attr),
full_matrix_projection(input=input_tmp[1], param_attr=lstm_para_attr)
]
)
lstm = lstmemory(name='lstm'+str(i),
input=mix_hidden,
act=ReluActivation(),
gate_act=SigmoidActivation(),
state_act=SigmoidActivation(),
reverse=((i % 2)==1),
bias_attr=std_0,
param_attr=lstm_para_attr)
input_tmp = [mix_hidden, lstm]
feature_out = mixed_layer(name='output',
size=label_dict_len,
bias_attr=std_default,
input=[full_matrix_projection(input=input_tmp[0], param_attr=hidden_para_attr),
full_matrix_projection(input=input_tmp[1], param_attr=lstm_para_attr)
],
)
lstm = lstmemory(
input=fc,
act=ReluActivation(),
reverse=(i % 2) == 1,
layer_attr=layer_attr)
input_tmp = [fc, lstm]
prob = fc_layer(
input=input_tmp,
size=label_dict_len,
act=SoftmaxActivation(),
param_attr=para_attr)
if not is_predict:
cls = classification_cost(input=prob, label=target)
outputs(cls)
crf_l = crf_layer( name = 'crf',
size = label_dict_len,
input = feature_out,
label = target,
param_attr=ParameterAttribute(name='crfw',initial_std=default_std, learning_rate=mix_hidden_lr)
)
crf_dec_l = crf_decoding_layer(name = 'crf_dec_l',
size = label_dict_len,
input = feature_out,
label = target,
param_attr=ParameterAttribute(name='crfw')
)
eval = sum_evaluator(input=crf_dec_l)
outputs(crf_l)
else:
outputs(prob)
crf_dec_l = crf_decoding_layer(name = 'crf_dec_l',
size = label_dict_len,
input = feature_out,
param_attr=ParameterAttribute(name='crfw')
)
outputs(crf_dec_l)
demo/semantic_role_labeling/predict.py
浏览文件 @ e4880016
......@@ -26,7 +26,7 @@ UNK_IDX = 0
class Prediction():
def __init__(self, train_conf, dict_file, model_dir, label_file):
def __init__(self, train_conf, dict_file, model_dir, label_file, predicate_dict_file):
"""
train_conf: trainer configure.
dict_file: word dictionary file name.
......@@ -35,26 +35,41 @@ class Prediction():
self.dict = {}
self.labels = {}
self.predicate_dict={}
self.labels_reverse = {}
self.load_dict_label(dict_file, label_file)
self.load_dict_label(dict_file, label_file, predicate_dict_file)
len_dict = len(self.dict)
len_label = len(self.labels)
conf = parse_config(train_conf, 'dict_len=' + str(len_dict) +
',label_len=' + str(len_label) + ',is_predict=True')
len_pred = len(self.predicate_dict)
conf = parse_config(
train_conf,
'dict_len=' + str(len_dict) +
',label_len=' + str(len_label) +
',pred_len=' + str(len_pred) +
',is_predict=True')
self.network = swig_paddle.GradientMachine.createFromConfigProto(
conf.model_config)
self.network.loadParameters(model_dir)
slots = [
integer_value_sequence(len_dict),
integer_value_sequence(len_pred),
integer_value_sequence(len_dict),
integer_value_sequence(len_dict),
integer_value_sequence(len_dict),
integer_value_sequence(len_dict),
integer_value_sequence(len_dict),
integer_value_sequence(2)
]
integer_value_sequence(len_dict), integer_value_sequence(len_dict),
integer_value_sequence(len_dict), integer_value_sequence(len_dict),
integer_value_sequence(len_dict), integer_value_sequence(2)
]
self.converter = DataProviderConverter(slots)
def load_dict_label(self, dict_file, label_file):
def load_dict_label(self, dict_file, label_file, predicate_dict_file):
"""
Load dictionary from self.dict_file.
"""
......@@ -65,39 +80,42 @@ class Prediction():
self.labels[line.strip()] = line_count
self.labels_reverse[line_count] = line.strip()
for line_count, line in enumerate(open(predicate_dict_file, 'r')):
self.predicate_dict[line.strip()] = line_count
def get_data(self, data_file):
"""
Get input data of paddle format.
"""
with open(data_file, 'r') as fdata:
for line in fdata:
sentence, predicate, ctx_n1, ctx_0, ctx_p1, mark, label = line.strip(
sentence, predicate, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2, mark, label = line.strip(
).split('\t')
words = sentence.split()
sen_len = len(words)
word_slot = [self.dict.get(w, UNK_IDX) for w in words]
predicate_slot = [self.dict.get(predicate, UNK_IDX)] * sen_len
predicate_slot = [self.predicate_dict.get(predicate, UNK_IDX)] * sen_len
ctx_n2_slot = [self.dict.get(ctx_n2, UNK_IDX)] * sen_len
ctx_n1_slot = [self.dict.get(ctx_n1, UNK_IDX)] * sen_len
ctx_0_slot = [self.dict.get(ctx_0, UNK_IDX)] * sen_len
ctx_p1_slot = [self.dict.get(ctx_p1, UNK_IDX)] * sen_len
ctx_p2_slot = [self.dict.get(ctx_p2, UNK_IDX)] * sen_len
marks = mark.split()
mark_slot = [int(w) for w in marks]
yield word_slot, predicate_slot, ctx_n2_slot, ctx_n1_slot, \
ctx_0_slot, ctx_p1_slot, ctx_p2_slot, mark_slot
yield word_slot, predicate_slot, ctx_n1_slot, \
ctx_0_slot, ctx_p1_slot, mark_slot
def predict(self, data_file):
def predict(self, data_file, output_file):
"""
data_file: file name of input data.
"""
input = self.converter(self.get_data(data_file))
output = self.network.forwardTest(input)
prob = output[0]["value"]
lab = list(np.argsort(-prob)[:, 0])
lab = output[0]["id"].tolist()
with open(data_file, 'r') as fin, open('predict.res', 'w') as fout:
with open(data_file, 'r') as fin, open(output_file, 'w') as fout:
index = 0
for line in fin:
sen = line.split('\t')[0]
......@@ -109,8 +127,8 @@ class Prediction():
def option_parser():
usage = ("python predict.py -c config -w model_dir "
"-d word dictionary -l label_file -i input_file")
usage = ("python predict.py -c config -w model_dir "
"-d word dictionary -l label_file -i input_file -p pred_dict_file")
parser = OptionParser(usage="usage: %s [options]" % usage)
parser.add_option(
"-c",
......@@ -131,6 +149,13 @@ def option_parser():
dest="label_file",
default=None,
help="label file")
parser.add_option(
"-p",
"--predict_dict_file",
action="store",
dest="predict_dict_file",
default=None,
help="predict_dict_file")
parser.add_option(
"-i",
"--data",
......@@ -144,6 +169,14 @@ def option_parser():
dest="model_path",
default=None,
help="model path")
parser.add_option(
"-o",
"--output_file",
action="store",
dest="output_file",
default=None,
help="output file")
return parser.parse_args()
......@@ -154,10 +187,12 @@ def main():
dict_file = options.dict_file
model_path = options.model_path
label_file = options.label_file
predict_dict_file = options.predict_dict_file
output_file = options.output_file
swig_paddle.initPaddle("--use_gpu=0")
predict = Prediction(train_conf, dict_file, model_path, label_file)
predict.predict(data_file)
predict = Prediction(train_conf, dict_file, model_path, label_file, predict_dict_file)
predict.predict(data_file,output_file)
if __name__ == '__main__':
......
demo/semantic_role_labeling/predict.sh
浏览文件 @ e4880016
......@@ -26,15 +26,18 @@ LOG=`get_best_pass $log`
LOG=(${LOG})
best_model_path="output/pass-${LOG[1]}"
config_file=db_lstm.py
dict_file=./data/src.dict
label_file=./data/tgt.dict
dict_file=./data/wordDict.txt
label_file=./data/targetDict.txt
predicate_dict_file=./data/verbDict.txt
input_file=./data/feature
output_file=predict.res
python predict.py \
-c $config_file \
-w $best_model_path \
-l $label_file \
-p $predicate_dict_file \
-d $dict_file \
-i $input_file
-i $input_file \
-o $output_file
demo/semantic_role_labeling/test.sh
浏览文件 @ e4880016
......@@ -36,4 +36,5 @@ paddle train \
--job=test \
--use_gpu=false \
--config_args=is_test=1 \
--test_all_data_in_one_period=1 \
2>&1 | tee 'test.log'
demo/semantic_role_labeling/train.sh
浏览文件 @ e4880016
......@@ -16,11 +16,14 @@
set -e
paddle train \
--config=./db_lstm.py \
--use_gpu=0 \
--log_period=5000 \
--trainer_count=1 \
--show_parameter_stats_period=5000 \
--save_dir=./output \
--trainer_count=4 \
--log_period=10 \
--num_passes=500 \
--use_gpu=false \
--show_parameter_stats_period=10 \
--num_passes=10000 \
--average_test_period=10000000 \
--init_model_path=./data \
--load_missing_parameter_strategy=rand \
--test_all_data_in_one_period=1 \
2>&1 | tee 'train.log'
2>&1 | tee 'train.log'
demo/sentiment/trainer_config.py
浏览文件 @ e4880016
......@@ -29,6 +29,7 @@ settings(
batch_size=128,
learning_rate=2e-3,
learning_method=AdamOptimizer(),
average_window=0.5,
regularization=L2Regularization(8e-4),
gradient_clipping_threshold=25)
......
doc/algorithm/rnn/rnn.rst
浏览文件 @ e4880016
......@@ -17,7 +17,7 @@ PaddlePaddle does not need any preprocessing to sequence data, such as padding.
.. code-block:: python
settings.slots = [
settings.input_types = [
integer_value_sequence(len(settings.src_dict)),
integer_value_sequence(len(settings.trg_dict)),
integer_value_sequence(len(settings.trg_dict))]
......
doc/demo/semantic_role_labeling/semantic_role_labeling.md
浏览文件 @ e4880016
# Semantic Role labeling Tutorial #
Semantic role labeling (SRL) is a form of shallow semantic parsing whose goal is to discover the predicate-argument structure of each predicate in a given input sentence. SRL is useful as an intermediate step in a wide range of natural language processing tasks, such as information extraction. automatic document categorization and question answering. An instance is as following [1]:
[ <sub>A0</sub> He ] [ <sub>AM-MOD</sub> would ][ <sub>AM-NEG</sub> n’t ] [ <sub>V</sub> accept] [ <sub>A1</sub> anything of value ] from [<sub>A2</sub> those he was writing about ].
- V: verb
- A0: acceptor
- A1: thing accepted
- A2: accepted-from
- A3: Attribute
- AM-MOD: modal
- AM-NEG: negation
Given the verb "accept", the chunks in sentence would play certain semantic roles. Here, the label scheme is from Penn Proposition Bank.
To this date, most of the successful SRL systems are built on top of some form of parsing results where pre-defined feature templates over the syntactic structure are used. This tutorial will present an end-to-end system using deep bidirectional long short-term memory (DB-LSTM)[2] for solving the SRL task, which largely outperforms the previous state-of-the-art systems. The system regards SRL task as the sequence labelling problem.
## Data Description
The relevant paper[2] takes the data set in CoNLL-2005&2012 Shared Task for training and testing. Accordingto data license, the demo adopts the test data set of CoNLL-2005, which can be reached on website.
To download and process the original data, user just need to execute the following command:
```bash
cd data
./get_data.sh
```
Several new files appear in the `data `directory as follows.
```bash
conll05st-release：the test data set of CoNll-2005 shared task
test.wsj.words：the Wall Street Journal data sentences
test.wsj.props: the propositional arguments
src.dict：the dictionary of words in sentences
tgt.dict：the labels dictionary
feature: the extracted features from data set
```
## Training
### DB-LSTM
Please refer to the Sentiment Analysis demo to learn more about the long short-term memory unit.
Unlike Bidirectional-LSTM that used in Sentiment Analysis demo, the DB-LSTM adopts another way to stack LSTM layer. First a standard LSTM processes the sequence in forward direction. The input and output of this LSTM layer are taken by the next LSTM layer as input, processed in reversed direction. These two standard LSTM layers compose a pair of LSTM. Then we stack LSTM layers pair after pair to obtain the deep LSTM model.
The following figure shows a temporal expanded 2-layer DB-LSTM network.
<center>
![pic](./network_arch.png)
</center>
### Features
Two input features play an essential role in this pipeline: predicate (pred) and argument (argu). Two other features: predicate context (ctx-p) and region mark (mr) are also adopted. Because a single predicate word can not exactly describe the predicate information, especially when the same words appear more than one times in a sentence. With the predicate context, the ambiguity can be largely eliminated. Similarly, we use region mark m<sub>r</sub> = 1 to denote the argument position if it locates in the predicate context region, or m<sub>r</sub> = 0 if does not. These four simple features are all we need for our SRL system. Features of one sample with context size set to 1 is showed as following[2]:
<center>
![pic](./feature.jpg)
</center>
In this sample, the coresponding labelled sentence is:
[ <sub>A1</sub> A record date ] has [ <sub>AM-NEG</sub> n't ] been [ <sub>V</sub> set ] .
In the demo, we adopt the feature template as above, consists of : `argument`, `predicate`, `ctx-p (p=-1,0,1)`, `mark` and use `B/I/O` scheme to label each argument. These features and labels are stored in `feature` file, and separated by `\t`.
### Data Provider
`dataprovider.py` is the python file to wrap data. `hook()` function is to define the data slots for network. The Six features and label are all IndexSlots.
```
def hook(settings, word_dict, label_dict, **kwargs):
settings.word_dict = word_dict
settings.label_dict = label_dict
#all inputs are integral and sequential type
settings.slots = [
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(2),
integer_value_sequence(len(label_dict))]
```
The corresponding data iterator is as following:
```
@provider(use_seq=True, init_hook=hook)
def process(obj, file_name):
with open(file_name, 'r') as fdata:
for line in fdata:
sentence, predicate, ctx_n1, ctx_0, ctx_p1, mark, label = line.strip().split('\t')
words = sentence.split()
sen_len = len(words)
word_slot = [obj.word_dict.get(w, UNK_IDX) for w in words]
predicate_slot = [obj.word_dict.get(predicate, UNK_IDX)] * sen_len
ctx_n1_slot = [obj.word_dict.get(ctx_n1, UNK_IDX) ] * sen_len
ctx_0_slot = [obj.word_dict.get(ctx_0, UNK_IDX) ] * sen_len
ctx_p1_slot = [obj.word_dict.get(ctx_p1, UNK_IDX) ] * sen_len
marks = mark.split()
mark_slot = [int(w) for w in marks]
label_list = label.split()
label_slot = [obj.label_dict.get(w) for w in label_list]
yield word_slot, predicate_slot, ctx_n1_slot, ctx_0_slot, ctx_p1_slot, mark_slot, label_slot
```
The `process`function yield 7 lists which are six features and labels.
### Neural Network Config
`db_lstm.py` is the neural network config file to load the dictionaries and define the data provider module and network architecture during the training procedure.
Seven `data_layer` load instances from data provider. Six features are transformed into embedddings respectively, and mixed by `mixed_layer` . Deep bidirectional LSTM layers extract features for the softmax layer. The objective function is cross entropy of labels.
### Run Training
The script for training is `train.sh`, user just need to execute:
```bash
./train.sh
```
The content in `train.sh`:
```
paddle train \
--config=./db_lstm.py \
--save_dir=./output \
--trainer_count=4 \
--log_period=10 \
--num_passes=500 \
--use_gpu=false \
--show_parameter_stats_period=10 \
--test_all_data_in_one_period=1 \
2>&1 | tee 'train.log'
```
- \--config=./db_lstm.py : network config file.
- \--save_di=./output: output path to save models.
- \--trainer_count=4 : set thread number (or GPU count).
- \--log_period=10 : print log every 20 batches.
- \--num_passes=500: set pass number, one pass in PaddlePaddle means training all samples in dataset one time.
- \--use_gpu=false: use CPU to train, set true, if you install GPU version of PaddlePaddle and want to use GPU to train.
- \--show_parameter_stats_period=10: show parameter statistic every 100 batches.
- \--test_all_data_in_one_period=1: test all data in every testing.
After training, the models will be saved in directory `output`.
### Run testing
The script for testing is `test.sh`, user just need to execute:
```bash
./test.sh
```
The main part in `tesh.sh`
```
paddle train \
--config=./db_lstm.py \
--model_list=$model_list \
--job=test \
--config_args=is_test=1 \
```
- \--config=./db_lstm.py: network config file
- \--model_list=$model_list.list: model list file
- \--job=test: indicate the test job
- \--config_args=is_test=1: flag to indicate test
### Run prediction
The script for prediction is `predict.sh`, user just need to execute:
```bash
./predict.sh
```
In `predict.sh`, user should offer the network config file, model path, label file, word dictionary file, feature file
```
python predict.py
-c $config_file
-w $model_path
-l $label_file
-d $dict_file
-i $input_file
```
`predict.py` is the main executable python script, which includes functions: load model, load data, data prediction. The network model will output the probability distribution of labels. In the demo, we take the label with maximum probability as result. User can also implement the beam search or viterbi decoding upon the probability distribution matrix.
After prediction, the result is saved in `predict.res`.
## Reference
[1] Martha Palmer, Dan Gildea, and Paul Kingsbury. The Proposition Bank: An Annotated Corpus of Semantic Roles , Computational Linguistics, 31(1), 2005.
[2] Zhou, Jie, and Wei Xu. "End-to-end learning of semantic role labeling using recurrent neural networks." Proceedings of the Annual Meeting of the Association for Computational Linguistics. 2015.
# Semantic Role labeling Tutorial #
Semantic role labeling (SRL) is a form of shallow semantic parsing whose goal is to discover the predicate-argument structure of each predicate in a given input sentence. SRL is useful as an intermediate step in a wide range of natural language processing tasks, such as information extraction. automatic document categorization and question answering. An instance is as following [1]:
[ <sub>A0</sub> He ] [ <sub>AM-MOD</sub> would ][ <sub>AM-NEG</sub> n’t ] [ <sub>V</sub> accept] [ <sub>A1</sub> anything of value ] from [<sub>A2</sub> those he was writing about ].
- V: verb
- A0: acceptor
- A1: thing accepted
- A2: accepted-from
- A3: Attribute
- AM-MOD: modal
- AM-NEG: negation
Given the verb "accept", the chunks in sentence would play certain semantic roles. Here, the label scheme is from Penn Proposition Bank.
To this date, most of the successful SRL systems are built on top of some form of parsing results where pre-defined feature templates over the syntactic structure are used. This tutorial will present an end-to-end system using deep bidirectional long short-term memory (DB-LSTM)[2] for solving the SRL task, which largely outperforms the previous state-of-the-art systems. The system regards SRL task as the sequence labelling problem.
## Data Description
The relevant paper[2] takes the data set in CoNLL-2005&2012 Shared Task for training and testing. Accordingto data license, the demo adopts the test data set of CoNLL-2005, which can be reached on website.
To download and process the original data, user just need to execute the following command:
```bash
cd data
./get_data.sh
```
Several new files appear in the `data `directory as follows.
```bash
conll05st-release：the test data set of CoNll-2005 shared task
test.wsj.words：the Wall Street Journal data sentences
test.wsj.props: the propositional arguments
feature: the extracted features from data set
```
## Training
### DB-LSTM
Please refer to the Sentiment Analysis demo to learn more about the long short-term memory unit.
Unlike Bidirectional-LSTM that used in Sentiment Analysis demo, the DB-LSTM adopts another way to stack LSTM layer. First a standard LSTM processes the sequence in forward direction. The input and output of this LSTM layer are taken by the next LSTM layer as input, processed in reversed direction. These two standard LSTM layers compose a pair of LSTM. Then we stack LSTM layers pair after pair to obtain the deep LSTM model.
The following figure shows a temporal expanded 2-layer DB-LSTM network.
<center>
![pic](./network_arch.png)
</center>
### Features
Two input features play an essential role in this pipeline: predicate (pred) and argument (argu). Two other features: predicate context (ctx-p) and region mark (mr) are also adopted. Because a single predicate word can not exactly describe the predicate information, especially when the same words appear more than one times in a sentence. With the predicate context, the ambiguity can be largely eliminated. Similarly, we use region mark m<sub>r</sub> = 1 to denote the argument position if it locates in the predicate context region, or m<sub>r</sub> = 0 if does not. These four simple features are all we need for our SRL system. Features of one sample with context size set to 1 is showed as following[2]:
<center>
![pic](./feature.jpg)
</center>
In this sample, the coresponding labelled sentence is:
[ <sub>A1</sub> A record date ] has [ <sub>AM-NEG</sub> n't ] been [ <sub>V</sub> set ] .
In the demo, we adopt the feature template as above, consists of : `argument`, `predicate`, `ctx-p (p=-1,0,1)`, `mark` and use `B/I/O` scheme to label each argument. These features and labels are stored in `feature` file, and separated by `\t`.
### Data Provider
`dataprovider.py` is the python file to wrap data. `hook()` function is to define the data slots for network. The Six features and label are all IndexSlots.
```
def hook(settings, word_dict, label_dict, **kwargs):
settings.word_dict = word_dict
settings.label_dict = label_dict
#all inputs are integral and sequential type
settings.slots = [
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(predicate_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(len(word_dict)),
integer_value_sequence(2),
integer_value_sequence(len(label_dict))]
```
The corresponding data iterator is as following:
```
@provider(init_hook=hook, should_shuffle=True, calc_batch_size=get_batch_size,
can_over_batch_size=False, cache=CacheType.CACHE_PASS_IN_MEM)
def process(settings, file_name):
with open(file_name, 'r') as fdata:
for line in fdata:
sentence, predicate, ctx_n2, ctx_n1, ctx_0, ctx_p1, ctx_p2, mark, label = \
line.strip().split('\t')
words = sentence.split()
sen_len = len(words)
word_slot = [settings.word_dict.get(w, UNK_IDX) for w in words]
predicate_slot = [settings.predicate_dict.get(predicate)] * sen_len
ctx_n2_slot = [settings.word_dict.get(ctx_n2, UNK_IDX)] * sen_len
ctx_n1_slot = [settings.word_dict.get(ctx_n1, UNK_IDX)] * sen_len
ctx_0_slot = [settings.word_dict.get(ctx_0, UNK_IDX)] * sen_len
ctx_p1_slot = [settings.word_dict.get(ctx_p1, UNK_IDX)] * sen_len
ctx_p2_slot = [settings.word_dict.get(ctx_p2, UNK_IDX)] * sen_len
marks = mark.split()
mark_slot = [int(w) for w in marks]
label_list = label.split()
label_slot = [settings.label_dict.get(w) for w in label_list]
yield word_slot, predicate_slot, ctx_n2_slot, ctx_n1_slot, \
ctx_0_slot, ctx_p1_slot, ctx_p2_slot, mark_slot, label_slot
```
The `process`function yield 9 lists which are 8 features and label.
### Neural Network Config
`db_lstm.py` is the neural network config file to load the dictionaries and define the data provider module and network architecture during the training procedure.
Nine `data_layer` load instances from data provider. Eight features are transformed into embedddings respectively, and mixed by `mixed_layer` . Deep bidirectional LSTM layers extract features for the softmax layer. The objective function is cross entropy of labels.
### Run Training
The script for training is `train.sh`, user just need to execute:
```bash
./train.sh
```
The content in `train.sh`:
```
paddle train \
--config=./db_lstm.py \
--use_gpu=0 \
--log_period=5000 \
--trainer_count=1 \
--show_parameter_stats_period=5000 \
--save_dir=./output \
--num_passes=10000 \
--average_test_period=10000000 \
--init_model_path=./data \
--load_missing_parameter_strategy=rand \
--test_all_data_in_one_period=1 \
2>&1 | tee 'train.log'
```
- \--config=./db_lstm.py : network config file.
- \--use_gpu=false: use CPU to train, set true, if you install GPU version of PaddlePaddle and want to use GPU to train, until now crf_layer do not support GPU
- \--log_period=500: print log every 20 batches.
- \--trainer_count=1: set thread number (or GPU count).
- \--show_parameter_stats_period=5000: show parameter statistic every 100 batches.
- \--save_dir=./output: output path to save models.
- \--num_passes=10000: set pass number, one pass in PaddlePaddle means training all samples in dataset one time.
- \--average_test_period=10000000: do test on average parameter every average_test_period batches
- \--init_model_path=./data: parameter initialization path
- \--load_missing_parameter_strategy=rand: random initialization unexisted parameters
- \--test_all_data_in_one_period=1: test all data in one period
After training, the models will be saved in directory `output`. Our training curve is as following:
<center>
![pic](./curve.jpg)
</center>
### Run testing
The script for testing is `test.sh`, user just need to execute:
```bash
./test.sh
```
The main part in `tesh.sh`
```
paddle train \
--config=./db_lstm.py \
--model_list=$model_list \
--job=test \
--config_args=is_test=1 \
```
- \--config=./db_lstm.py: network config file
- \--model_list=$model_list.list: model list file
- \--job=test: indicate the test job
- \--config_args=is_test=1: flag to indicate test
- \--test_all_data_in_one_period=1: test all data in 1 period
### Run prediction
The script for prediction is `predict.sh`, user just need to execute:
```bash
./predict.sh
```
In `predict.sh`, user should offer the network config file, model path, label file, word dictionary file, feature file
```
python predict.py
-c $config_file \
-w $best_model_path \
-l $label_file \
-p $predicate_dict_file \
-d $dict_file \
-i $input_file \
-o $output_file
```
`predict.py` is the main executable python script, which includes functions: load model, load data, data prediction. The network model will output the probability distribution of labels. In the demo, we take the label with maximum probability as result. User can also implement the beam search or viterbi decoding upon the probability distribution matrix.
After prediction, the result is saved in `predict.res`.
## Reference
[1] Martha Palmer, Dan Gildea, and Paul Kingsbury. The Proposition Bank: An Annotated Corpus of Semantic Roles , Computational Linguistics, 31(1), 2005.
[2] Zhou, Jie, and Wei Xu. "End-to-end learning of semantic role labeling using recurrent neural networks." Proceedings of the Annual Meeting of the Association for Computational Linguistics. 2015.
doc/demo/sentiment_analysis/sentiment_analysis.md
浏览文件 @ e4880016
......@@ -6,7 +6,7 @@ Sentiment analysis is also used to monitor social media based on large amount of
On the other hand, grabbing the user comments of products and analyzing their sentiment are useful to understand user preferences for companies, products, even competing products.
This tutorial will guide you through the process of training a Long Short Term Memory (LSTM) Network to classify the sentiment of sentences from [Large Movie Review Dataset](http://ai.stanford.edu/~amaas/data/sentiment/), sometimes known as the [Internet Movie Database (IMDB)](http://ai.stanford.edu/~amaas/papers/wvSent_acl2011.pdf). This dataset contains movie reviews along with their associated binary sentiment polarity labels, namely positive and negative. So randomly guessing yields 50% accuracy.
This tutorial will guide you through the process of training a Long Short Term Memory (LSTM) Network to classify the sentiment of sentences from [Large Movie Review Dataset](http://ai.stanford.edu/~amaas/data/sentiment/), sometimes known as the Internet Movie Database (IMDB). This dataset contains movie reviews along with their associated binary sentiment polarity labels, namely positive and negative. So randomly guessing yields 50% accuracy.
## Data Preparation
......@@ -39,7 +39,7 @@ imdbEr.txt imdb.vocab README test train
* imdbEr.txt: expected rating for each token in imdb.vocab.
* README: data documentation.
Both train and test set directory contains:
The file in train set directory is as follows. The test set also contains them except `unsup` and `urls_unsup.txt`.
```
labeledBow.feat neg pos unsup unsupBow.feat urls_neg.txt urls_pos.txt urls_unsup.txt
......@@ -151,6 +151,7 @@ settings(
batch_size=128,
learning_rate=2e-3,
learning_method=AdamOptimizer(),
average_window=0.5,
regularization=L2Regularization(8e-4),
gradient_clipping_threshold=25
)
......@@ -163,17 +164,18 @@ stacked_lstm_net(dict_dim, class_dim=class_dim,
* **Data Definition**:
* get\_config\_arg(): get arguments setted by `--config_args=xx` in commandline argument.
* Define TrainData and TestData provider, here using Python interface (PyDataProviderWrapper) of PaddlePaddle to load data. For details, you can refer to the document of PyDataProvider.
* Define data provider, here using Python interface to load data. For details, you can refer to the document of PyDataProvider2.
* **Algorithm Configuration**:
* use sgd algorithm.
* use adam optimization.
* set batch size of 128.
* set average sgd window.
* set global learning rate.
* use adam optimization.
* set average sgd window.
* set L2 regularization.
* set gradient clipping threshold.
* **Network Configuration**:
* dict_dim: get dictionary dimension.
* class_dim: set category number, IMDB has two label, namely positive and negative label.
* dict_dim: dictionary dimension.
* class_dim: category number, IMDB has two label, namely positive and negative label.
* `stacked_lstm_net`: predefined network as shown in Figure 3, use this network by default.
* `bidirectional_lstm_net`: predefined network as shown in Figure 2.
......
doc/dev/new_layer/new_layer.rst
浏览文件 @ e4880016
......@@ -60,7 +60,7 @@ Implement C++ Class
The C++ class of the layer implements the initialization, forward, and backward part of the layer. The fully connected layer is at :code:`paddle/gserver/layers/FullyConnectedLayer.h` and :code:`paddle/gserver/layers/FullyConnectedLayer.cpp`. We list simplified version of the code below.
It needs to derive the base class :code:`paddle::BaseLayer`, and it needs to override the following functions:
It needs to derive the base class :code:`paddle::Layer`, and it needs to override the following functions:
- constructor and destructor.
- :code:`init` function. It is used to initialize the parameters and settings.
......
doc/optimization/gpu_profiling.rst
浏览文件 @ e4880016
......@@ -53,7 +53,7 @@ above profilers.
.. literalinclude:: ../../paddle/math/tests/test_GpuProfiler.cpp
:language: c++
:lines: 107-121
:lines: 111-124
:linenos:
The above code snippet includes two methods, you can use any of them to profile the regions of interest.
......@@ -75,12 +75,12 @@ To enable built-in timer in PaddlePaddle, first you have to add :code:`REGISTER_
Then, all information could be stamped in the console via :code:`printStatus` or :code:`printAllStatus` function.
As a simple example, consider the following:
1. Add :code:`REGISTER_TIMER_INFO` and :code:`printStatus` functions (see the emphasize-lines).
1. Add :code:`REGISTER_TIMER_INFO` and :code:`printAllStatus` functions (see the emphasize-lines).
.. literalinclude:: ../../paddle/math/tests/test_GpuProfiler.cpp
:language: c++
:lines: 107-121
:emphasize-lines: 10-11,14
:lines: 111-124
:emphasize-lines: 8-10,13
:linenos:
2. Configure cmake with **WITH_TIMER** and recompile PaddlePaddle.
......@@ -126,8 +126,8 @@ To use this command line profiler **nvprof**, you can simply issue the following
.. literalinclude:: ../../paddle/math/tests/test_GpuProfiler.cpp
:language: c++
:lines: 107-121
:emphasize-lines: 7-8
:lines: 111-124
:emphasize-lines: 6-7
:linenos:
2. Configure cmake with **WITH_PROFILER** and recompile PaddlePaddle.
......
doc/source/api/api.rst → doc/source/api.rst
浏览文件 @ e4880016
API
========
===
.. doxygenfile:: paddle/api/PaddleAPI.h
.. doxygenfile:: paddle/api/Internal.h
doc/source/cuda/cuda/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
CUDA
====================
.. toctree::
:maxdepth: 3
cuda.rst
doc/source/cuda/matrix/index.rst → doc/source/cuda/index.rst
浏览文件 @ e4880016
Matrix
====================
CUDA
====
.. toctree::
:maxdepth: 3
:maxdepth: 2
matrix.rst
nn.rst
utils.rst
doc/source/cuda/matrix/matrix.rst → doc/source/cuda/matrix.rst
浏览文件 @ e4880016
Matrix
=======
======
Base Matrix
-------------
Base
----
hl_matrix.h
``````````````````
```````````
.. doxygenfile:: paddle/cuda/include/hl_matrix.h
hl_matrix_base.h
``````````````````
````````````````
.. doxygenfile:: paddle/cuda/include/hl_matrix_base.cuh
hl_matrix_apply.cuh
``````````````````````
```````````````````
.. doxygenfile:: paddle/cuda/include/hl_matrix_apply.cuh
hl_matrix_ops.cuh
``````````````````````
`````````````````
.. doxygenfile:: paddle/cuda/include/hl_matrix_ops.cuh
hl_matrix_type.cuh
``````````````````````
``````````````````
.. doxygenfile:: paddle/cuda/include/hl_matrix_type.cuh
hl_sse_matrix_kernel.cuh
``````````````````````````
````````````````````````
.. doxygenfile:: paddle/cuda/include/hl_sse_matrix_kernel.cuh
Matrix Function
---------------
hl_batch_transpose.h
``````````````````````````
````````````````````
.. doxygenfile:: paddle/cuda/include/hl_batch_transpose.h
Sparse Matrix
--------------
hl_sparse.h
``````````````````
.. doxygenfile:: paddle/cuda/include/hl_sparse.h
hl_sparse.ph
``````````````````````
.. doxygenfile:: paddle/cuda/include/hl_sparse.ph
Others
---------------
hl_aggregate.h
``````````````````
``````````````
.. doxygenfile:: paddle/cuda/include/hl_aggregate.h
hl_top_k.h
``````````
.. doxygenfile:: paddle/cuda/include/hl_top_k.h
hl_table_apply.h
``````````````````
````````````````
.. doxygenfile:: paddle/cuda/include/hl_table_apply.h
hl_top_k.h
``````````````````
.. doxygenfile:: paddle/cuda/include/hl_top_k.h
Sparse Matrix
-------------
hl_sparse.h
```````````
.. doxygenfile:: paddle/cuda/include/hl_sparse.h
hl_sparse.ph
````````````
.. doxygenfile:: paddle/cuda/include/hl_sparse.ph
doc/source/cuda/rnn/rnn.rst → doc/source/cuda/nn.rst
浏览文件 @ e4880016
Neural Networks
==================
Neural Network
==============
Base
-------
----
.. doxygenfile:: paddle/cuda/include/hl_gpu.h
.. doxygenfile:: paddle/cuda/include/hl_cnn.h
.. doxygenfile:: paddle/cuda/include/hl_functions.h
.. doxygenfile:: paddle/cuda/include/hl_avx_functions.h
.. doxygenfile:: paddle/cuda/include/hl_device_functions.cuh
.. doxygenfile:: paddle/cuda/include/hl_gpu_functions.cuh
Activation Functions
-----------------------
.. doxygenfile:: paddle/cuda/include/hl_activation_functions.h
CNN Related APIs
----------------
.. doxygenfile:: paddle/cuda/include/hl_cnn.h
.. doxygenfile:: paddle/cuda/include/hl_cuda_cudnn.h
.. doxygenfile:: paddle/cuda/include/hl_cuda_cudnn.ph
RNN Related APIs
-----------------
----------------
.. doxygenfile:: paddle/cuda/include/hl_recurrent_apply.cuh
.. doxygenfile:: paddle/cuda/include/hl_sequence.h
LSTM Model
``````````````
``````````
.. doxygenfile:: paddle/cuda/include/hl_lstm.h
.. dpxygenfile:: paddle/cuda/include/hl_cpu_lstm.cuh
.. doxygenfile:: paddle/cuda/include/hl_gpu_lstm.cuh
.. doxygenfile:: paddle/cuda/include/hl_lstm_ops.cuh
GRU Model
````````````````
`````````
.. doxygenfile:: paddle/cuda/include/hl_gru_ops.cuh
.. doxygenfile:: paddle/cuda/include/hl_cpu_gru.cuh
.. doxygenfile:: paddle/cuda/include/hl_gpu_gru.cuh
doc/source/cuda/rnn/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
RNN
====================
.. toctree::
:maxdepth: 3
rnn.rst
doc/source/cuda/cuda/cuda.rst → doc/source/cuda/utils.rst
浏览文件 @ e4880016
Cuda
=============
Utils
=====
Dynamic Link Libs
--------------------------
hl_dso_loader.h
``````````````````
-----------------
.. doxygenfile:: paddle/cuda/include/hl_dso_loader.h
GPU Resources
----------------
-------------
hl_cuda.ph
``````````````
``````````
.. doxygenfile:: paddle/cuda/include/hl_cuda.ph
hl_cuda.h
``````````````
`````````
.. doxygenfile:: paddle/cuda/include/hl_cuda.h
CUDA Wrapper
--------------
HPPL Base
---------
.. doxygenfile:: paddle/cuda/include/hl_base.h
hl_cuda_cublas.h
``````````````````````
CUBLAS Wrapper
--------------
.. doxygenfile:: paddle/cuda/include/hl_cuda_cublas.h
hl_cuda_cudnn.h
``````````````````````
.. doxygenfile:: paddle/cuda/include/hl_cuda_cudnn.h
hl_cuda_cudnn.h
``````````````````````
.. doxygenfile:: paddle/cuda/include/hl_cuda_cudnn.ph
Timer
-----
.. doxygenfile:: paddle/cuda/include/hl_time.h
Thread Resource
---------------
.. doxygenfile:: paddle/cuda/include/hl_thread.ph
Device Function
---------------
.. doxygenfile:: paddle/cuda/include/hl_device_functions.cuh
doc/source/cuda/utils/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Utils
====================
.. toctree::
:maxdepth: 3
utils.rst
doc/source/cuda/utils/utils.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Utilities
===========
HPPL Base
------------
hl_base.h
``````````````
.. doxygenfile:: paddle/cuda/include/hl_base.h
Timer
-----------
hl_time.h
``````````````
.. doxygenfile:: paddle/cuda/include/hl_time.h
Thread Resource
-----------
hl_thread.ph
``````````````
.. doxygenfile:: paddle/cuda/include/hl_thread.ph
doc/source/gserver/activations/index.rst → doc/source/gserver/activations.rst
浏览文件 @ e4880016
Activations
=============
===========
.. doxygenclass:: paddle::ActivationFunction
:members:
doc/source/gserver/dataprovider/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Data Providers Documents
==========================
.. toctree::
:maxdepth: 3
dataproviders.rst
doc/source/gserver/dataprovider/dataproviders.rst → doc/source/gserver/dataproviders.rst
浏览文件 @ e4880016
==============
Data Providers
================
==============
Base DataProvider
------------------
DataProviders
=============
Base
----
.. doxygenclass:: paddle::DataProvider
:members:
DataProviderGroup
-------------------
-----------------
.. doxygenclass:: paddle::DataProviderGroup
:members:
MultiDataProvider
-------------------
-----------------
.. doxygenclass:: paddle::MultiDataProvider
:members:
PyDataProvider
===================
==============
IFieldScanner
-------------
......@@ -45,7 +49,7 @@ SparseValueScanner
:members:
SequenceScanner
------------------
---------------
.. doxygenclass:: paddle::SparseValueScanner
:members:
......@@ -69,8 +73,8 @@ IPyDataProvider
.. doxygenclass:: paddle::PyDataProvider2
:members:
Proto Data Provider
===================
ProtoDataProvider
=================
ProtoDataProvider
----------------
......@@ -78,6 +82,6 @@ ProtoDataProvider
:members:
ProtoSequenceDataProvider
----------------
-------------------------
.. doxygenclass:: paddle::ProtoSequenceDataProvider
:members:
doc/source/gserver/evaluators/evaluators.rst → doc/source/gserver/evaluators.rst
浏览文件 @ e4880016
Base Evaluator
==============
==========
Evaluators
==========
Base
====
Evaluator
---------
.. doxygenclass:: paddle::Evaluator
:members:
Utils
=====
Sum
===
SumEvaluator
------------
......
doc/source/gserver/evaluators/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Evaluators
==========
.. toctree::
:maxdepth: 3
evaluators.rst
doc/source/gserver/gradientmachines/gradientmachines.rst → doc/source/gserver/gradientmachines.rst
浏览文件 @ e4880016
Gradient Machines
================
=================
GradientMachine
---------------------
---------------
.. doxygenclass:: paddle::GradientMachine
:members:
GradientMachineModel
--------------------
GradientMachineMode
-------------------
.. doxygenclass:: paddle::IGradientMachineMode
:members:
MultiGradientMachine
---------------------
--------------------
.. doxygenclass:: paddle::MultiGradientMachine
:members:
......@@ -21,20 +21,7 @@ TrainerThread
.. doxygenclass:: paddle::TrainerThread
:members:
Recurrent Gradient Machines
---------------------------
RecurrentGradientMachine
------------------------
.. doxygenclass:: paddle::RecurrentGradientMachine
:members:
Networks
========
NeuralNetwork
-------------
.. doxygenclass:: paddle::NeuralNetwork
:members:
ParallelNeuralNetwork
---------------------
.. doxygenclass:: paddle::ParallelNeuralNetwork
:members:
doc/source/gserver/gradientmachines/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Gradient Machines Documents
=============================
.. toctree::
:maxdepth: 3
gradientmachines.rst
doc/source/gserver/index.rst 0 → 100644
浏览文件 @ e4880016
GServer
=======
.. toctree::
:maxdepth: 2
activations.rst
dataproviders.rst
evaluators.rst
gradientmachines.rst
layers.rst
neworks.rst
doc/source/gserver/layers/layer.rst → doc/source/gserver/layers.rst
浏览文件 @ e4880016
Base
======
Layers
======
Base
====
Layer
-----
.. doxygenclass:: paddle::Layer
......@@ -17,7 +21,7 @@ Operator
:members:
Data Layer
===========
==========
.. doxygenclass:: paddle::DataLayer
:members:
......@@ -58,6 +62,11 @@ CudnnConvLayer
.. doxygenclass:: paddle::CudnnConvLayer
:members:
ExpandConvBaseLayer
-------------------
.. doxygenclass:: paddle::ExpandConvBaseLayer
:members:
ExpandConvLayer
---------------
.. doxygenclass:: paddle::ExpandConvLayer
......@@ -86,6 +95,16 @@ CudnnPoolLayer
.. doxygenclass:: paddle::CudnnPoolLayer
:members:
SpatialPyramidPoolLayer
-----------------------
.. doxygenclass:: paddle::SpatialPyramidPoolLayer
:members:
MaxOutLayer
-----------
.. doxygenclass:: paddle::MaxOutLayer
:members:
Norm Layers
===========
......@@ -402,6 +421,11 @@ TransLayer
Sampling Layers
===============
BilinearInterpLayer
-------------------
.. doxygenclass:: paddle::BilinearInterpLayer
:members:
MultinomialSampler
------------------
.. doxygenclass:: paddle::MultinomialSampler
......
doc/source/gserver/layers/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Layers Documents
====================
.. toctree::
:maxdepth: 3
layer.rst
doc/source/gserver/neworks.rst 0 → 100644
浏览文件 @ e4880016
Networks
========
NeuralNetwork
-------------
.. doxygenclass:: paddle::NeuralNetwork
:members:
ParallelNeuralNetwork
---------------------
.. doxygenclass:: paddle::ParallelNeuralNetwork
:members:
doc/source/index.md 已删除 100644 → 0
浏览文件 @ 8393c19c
# Source Code Documents
## cuda
- [CUDA](cuda/cuda/index.rst)
- [Matrix](cuda/matrix/index.rst)
- [RNN](cuda/rnn/index.rst)
- [Utils](cuda/utils/index.rst)
## gserver
- [Activations](gserver/activations/index.rst)
- [Data Providers](gserver/dataprovider/index.rst)
- [Evaluators](gserver/evaluators/index.rst)
- [Gradient Machines](gserver/gradientmachines/index.rst)
- [Layers](gserver/layers/index.rst)
## math
- [Matrix](math/matrix/index.rst)
- [Utils](math/utils/index.rst)
## parameter
- [Parameter](parameter/parameter/index.rst)
- [Update](parameter/update/index.rst)
- [Optimizer](parameter/optimizer/index.rst)
## pserver
- [Client](pserver/client/index.rst)
- [Network](pserver/network/index.rst)
- [Server](pserver/server/index.rst)
## trainer
- [Trainer](trainer/trainer.rst)
## api
- [API](api/api.rst)
## utils
- [CustomStackTrace](utils/customStackTrace.rst)
- [Enumeration wrapper](utils/enum.rst)
- [Lock](utils/lock.rst)
- [Queue](utils/queue.rst)
- [Thread](utils/thread.rst)
doc/source/index.rst 0 → 100644
浏览文件 @ e4880016
Source Code Documents
=====================
.. toctree::
:maxdepth: 1
gserver/index.rst
trainer.rst
parameter/index.rst
pserver/index.rst
api.rst
cuda/index.rst
math/index.rst
utils/index.rst
doc/source/math/functions.rst 0 → 100644
浏览文件 @ e4880016
Functions
=========
MathFunctions
-------------
.. doxygenfile:: paddle/math/MathFunctions.h
SIMDFunctions
-------------
.. doxygenfile:: paddle/math/SIMDFunctions.h
doc/source/math/index.rst 0 → 100644
浏览文件 @ e4880016
Math
====
.. toctree::
:maxdepth: 2
vector.rst
matrix.rst
functions.rst
utils.rst
doc/source/math/matrix.rst 0 → 100644
浏览文件 @ e4880016
Matrix
======
Base
----
BaseMatrix Template
```````````````````
.. doxygenclass:: paddle::BaseMatrixT
:members:
Matrix
``````
.. doxygenclass:: paddle::Matrix
:members:
MatrixOffset
````````````
.. doxygenclass:: paddle::MatrixOffset
:members:
CpuMatrix
---------
CpuMatrix
`````````
.. doxygenclass:: paddle::CpuMatrix
:members:
SharedCpuMatrix
```````````````
.. doxygenclass:: paddle::SharedCpuMatrix
:members:
GpuMatrix
---------
.. doxygenclass:: paddle::GpuMatrix
:members:
CpuSparseMatrix
---------------
CpuSparseMatrix
```````````````
.. doxygenclass:: paddle::CpuSparseMatrix
:members:
SparseRowCpuMatrix
``````````````````
.. doxygenclass:: paddle::SparseRowCpuMatrix
:members:
SparseAutoGrowRowCpuMatrix
``````````````````````````
.. doxygenclass:: paddle::SparseAutoGrowRowCpuMatrix
:members:
SparsePrefetchRowCpuMatrix
``````````````````````````
.. doxygenclass:: paddle::SparsePrefetchRowCpuMatrix
:members:
SparseRowIdsCpuMatrix
`````````````````````
.. doxygenclass:: paddle::SparseRowIdsCpuMatrix
:members:
CacheRowCpuMatrix
`````````````````
.. doxygenclass:: paddle::CacheRowCpuMatrix
:members:
GpuSparseMatrix
---------------
.. doxygenclass:: paddle::GpuSparseMatrix
:members:
doc/source/math/matrix/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Matrix Documents
====================
.. toctree::
:maxdepth: 3
matrix.rst
doc/source/math/matrix/matrix.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Matrix
=======
Base
--------
.. doxygenfile:: paddle/math/BaseMatrix.h
Sparse Matrix
----------------
.. doxygenfile:: paddle/math/Matrix.h
.. doxygenfile:: paddle/math/Vector.h
.. doxygenfile:: paddle/math/MathUtils.h
.. doxygenfile:: paddle/math/SparseMatrix.h
.. doxygenfile:: paddle/math/SparseRowMatrix.h
.. doxygenfile:: paddle/math/CpuSparseMatrix.h
Others
----------
.. doxygenfile:: paddle/math/MathFunctions.h
.. doxygenfile:: paddle/math/SIMDFunctions.h
doc/source/math/utils/utils.rst → doc/source/math/utils.rst
浏览文件 @ e4880016
Utils
=======
Memory Manager
==============
Memory Handle
--------------
-------------
.. doxygenfile:: paddle/math/MemoryHandle.h
Allocator
---------
.. doxygenfile:: paddle/math/Allocator.h
PoolAllocator
`````````````
.. doxygenfile:: paddle/math/PoolAllocator.h
Storage
-------
.. doxygenfile:: paddle/math/Storage.h
doc/source/math/utils/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Utils Documents
====================
.. toctree::
:maxdepth: 3
utils.rst
doc/source/math/vector.rst 0 → 100644
浏览文件 @ e4880016
Vector
======
BaseVector
``````````
.. doxygenclass:: paddle::BaseVector
:members:
Vector Template
```````````````
.. doxygenclass:: paddle::VectorT
:members:
CpuVector Template
``````````````````
.. doxygenclass:: paddle::CpuVectorT
:members:
GpuVector Template
``````````````````
.. doxygenclass:: paddle::GpuVectorT
:members:
ParallelCpuVector Template
``````````````````````````
.. doxygenclass:: paddle::ParallelCpuVectorT
:members:
ParallelGpuVector Template
``````````````````````````
.. doxygenclass:: paddle::ParallelGpuVectorT
:members:
CpuGpuVector Template
`````````````````````
.. doxygenclass:: paddle::CpuGpuVectorT
:members:
doc/source/parameter/optimizer/index.rst → doc/source/parameter/index.rst
浏览文件 @ e4880016
Parameter Documents
====================
Parameter
=========
.. toctree::
:maxdepth: 3
:maxdepth: 2
parameter.rst
optimizer.rst
updater.rst
doc/source/parameter/optimizer/optimizer.rst → doc/source/parameter/optimizer.rst
浏览文件 @ e4880016
Optimizer
============
=========
ParameterOptimizer
------------------
.. doxygenfile:: paddle/parameter/ParameterOptimizer.h
Regularizer
-----------
.. doxygenfile:: paddle/parameter/Regularizer.h
FirstOrderOptimizer
-------------------
.. doxygenfile:: paddle/parameter/FirstOrderOptimizer.h
AverageOptimizer
----------------
.. doxygenfile:: paddle/parameter/AverageOptimizer.h
.. doxygenfile:: paddle/parameter/ParameterOptimizer.h
OptimizerWithRegularizer
------------------------
.. doxygenfile:: paddle/parameter/OptimizerWithRegularizer.h
doc/source/parameter/parameter/parameter.rst → doc/source/parameter/parameter.rst
浏览文件 @ e4880016
Parameter
=============
Weight
--------
.. doxygenfile:: paddle/parameter/Weight.h
Regularizer
------------
.. doxygenfile:: paddle/parameter/Regularizer.h
=========
Parameter
-------------
---------
.. doxygenfile:: paddle/parameter/Argument.h
.. doxygenfile:: paddle/parameter/Parameter.h
.. doxygenfile:: paddle/parameter/ParallelParameter.h
Weight
------
.. doxygenfile:: paddle/parameter/Weight.h
doc/source/parameter/parameter/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Parameter Documents
====================
.. toctree::
:maxdepth: 3
parameter.rst
doc/source/parameter/update/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Parameter Documents
====================
.. toctree::
:maxdepth: 3
update.rst
doc/source/parameter/update/update.rst → doc/source/parameter/updater.rst
浏览文件 @ e4880016
Update
==========
Updater
=======
Base
----
.. doxygenfile:: paddle/parameter/ParameterUpdaterBase.h
Hook
----
.. doxygenfile:: paddle/parameter/ParameterUpdaterHook.h
.. doxygenfile:: paddle/parameter/ParameterUpdateFunctions.h
Functions
---------
.. doxygenfile:: paddle/parameter/ParameterUpdateFunctions.h
doc/source/pserver/client.rst 0 → 100644
浏览文件 @ e4880016
Client
======
BaseClient
----------
.. doxygenclass:: paddle::BaseClient
:members:
ParameterClient2
----------------
.. doxygenclass:: paddle::ParameterClient2
:members:
doc/source/pserver/client/client.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Client
=========
.. doxygenclass:: paddle::BaseClient
:members:
:protected-members:
:private-members:
:undoc-members:
.. doxygenclass:: paddle::ParameterClient2
:members:
:protected-members:
:private-members:
:undoc-members:
doc/source/pserver/client/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Client Documents
====================
.. toctree::
:maxdepth: 3
client.rst
doc/source/pserver/index.rst 0 → 100644
浏览文件 @ e4880016
PServer
=======
.. toctree::
:maxdepth: 2
client.rst
network.rst
server.rst
utils.rst
doc/source/pserver/network.rst 0 → 100644
浏览文件 @ e4880016
Network
=======
SocketServer
------------
.. doxygenclass:: paddle::SocketServer
:members:
SocketWorker
------------
.. doxygenclass:: paddle::SocketWorker
:members:
SocketClient
------------
.. doxygenclass:: paddle::SocketClient
:members:
SocketChannel
-------------
.. doxygenclass:: paddle::SocketChannel
:members:
MessageReader
-------------
.. doxygenclass:: paddle::MsgReader
:members:
doc/source/pserver/network/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Network Documents
====================
.. toctree::
:maxdepth: 3
network.rst
doc/source/pserver/network/network.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Network
==========
Socket Server
----------------
.. doxygenclass:: paddle::SocketServer
:members:
:protected-members:
:private-members:
:undoc-members:
Socket Worker
----------------
.. doxygenclass:: paddle::SocketWorker
:members:
:protected-members:
:private-members:
:undoc-members:
Socket Client
----------------
.. doxygenclass:: paddle::SocketClient
:members:
:protected-members:
:private-members:
:undoc-members:
Socket Channel
---------------
.. doxygenclass:: paddle::SocketChannel
:members:
:protected-members:
:private-members:
:undoc-members:
Message Reader
---------------
.. doxygenclass:: paddle::MsgReader
:members:
:protected-members:
:private-members:
:undoc-members:
doc/source/pserver/server.rst 0 → 100644
浏览文件 @ e4880016
Server
======
ProtoServer
-----------
.. doxygenclass:: paddle::ProtoServer
:members:
ParameterServer2
----------------
.. doxygenclass:: paddle::ParameterServer2
:members:
doc/source/pserver/server/index.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Server Documents
====================
.. toctree::
:maxdepth: 3
server.rst
doc/source/pserver/server/server.rst 已删除 100644 → 0
浏览文件 @ 8393c19c
Server
==========
.. doxygenclass:: paddle::ProtoServer
:members:
:protected-members:
:private-members:
:undoc-members:
.. doxygenclass:: paddle::ParameterServer2
:members:
:protected-members:
:private-members:
:undoc-members:
doc/source/trainer/trainer.rst → doc/source/trainer.rst
浏览文件 @ e4880016
......@@ -14,7 +14,7 @@ RemoteParameterUpdater
:members:
ConcurrentRemoteParameterUpdater
---------------------------------
--------------------------------
.. doxygenclass:: paddle::ConcurrentRemoteParameterUpdater
:members:
......
doc/source/utils/customStackTrace.rst
浏览文件 @ e4880016
CustomStackTrace
================
class CustomStackTrace
----------------------
.. doxygenclass:: paddle::CustomStackTrace
:members:
doc/source/utils/enum.rst
浏览文件 @ e4880016
enumeration_wrapper
Enumeration wrapper
===================
namespace paddle::enumeration_wrapper
-------------------------------------
.. doxygennamespace:: paddle::enumeration_wrapper
doc/source/utils/index.rst 0 → 100644
浏览文件 @ e4880016
Utils
=====
.. toctree::
:maxdepth: 2
lock.rst
queue.rst
thread.rst
customStackTrace.rst
enum.rst
doc/source/utils/lock.rst
浏览文件 @ e4880016
Thread
======
Lock
====
class Thread
------------
.. doxygenclass:: paddle::Thread
RWLock
------
.. doxygenclass:: paddle::RWLock
:members:
class ThreadWorker
------------------
.. doxygenclass:: paddle::ThreadWorker
ReadLockGuard
-------------
.. doxygenclass:: paddle::ReadLockGuard
:members:
class SyncThreadPool
--------------------
.. doxygenclass:: paddle::SyncThreadPool
SpinLock
--------
.. doxygenclass:: paddle::SpinLock
:members:
class MultiThreadWorker
-----------------------
.. doxygenclass:: paddle::MultiThreadWorker
Semaphore
---------
.. doxygenclass:: paddle::Semaphore
:members:
class AsyncThreadPool
---------------------
ThreadBarrier
-------------
.. doxygenclass:: paddle::ThreadBarrier
:members:
.. doxygenclass:: paddle::AsyncThreadPool
LockedCondition
---------------
.. doxygenclass:: paddle::LockedCondition
:members:
doc/source/utils/queue.rst
浏览文件 @ e4880016
Queue
=====
class Queue
------------
Queue
-----
.. doxygenclass:: paddle::Queue
:members:
class BlockingQueue
-------------------
BlockingQueue
-------------
.. doxygenclass:: paddle::BlockingQueue
:members:
doc/source/utils/thread.rst
浏览文件 @ e4880016
Lock
====
Thread
======
class RWLock
------------
.. doxygenclass:: paddle::RWLock
Thread
------
.. doxygenclass:: paddle::Thread
:members:
class ReadLockGuard
-------------------
.. doxygenclass:: paddle::ReadLockGuard
ThreadWorker
------------
.. doxygenclass:: paddle::ThreadWorker
:members:
class SpinLock
SyncThreadPool
--------------
.. doxygenclass:: paddle::SpinLock
.. doxygenclass:: paddle::SyncThreadPool
:members:
class Semaphore
---------------
.. doxygenclass:: paddle::Semaphore
:members:
class ThreadBarrier
-------------------
.. doxygenclass:: paddle::ThreadBarrier
MultiThreadWorker
-----------------
.. doxygenclass:: paddle::MultiThreadWorker
:members:
class LockedCondition
---------------------
.. doxygenclass:: paddle::LockedCondition
AsyncThreadPool
---------------
.. doxygenclass:: paddle::AsyncThreadPool
:members:
doc_cn/build_and_install/cmake/cblas_settings.csv
浏览文件 @ e4880016
MKL_ROOT,mkl的路径，在${MKL_ROOT}/include下需要包含mkl.h，在${MKL_ROOT}/lib目录下需要包含 mkl_core，mkl_sequential和mkl_intel_lp64三个库
ATLAS_ROOT,ATLAS库的路径，在${ATLAS_ROOT}/include下需要包含cblas.h，而在${ATLAS_ROOT}/lib下需要包含cblas和atlas两个库
OPENBLAS_ROOT,在${OPENBLAS_ROOT}/include下需要包含cblas.h，而在${OPENBLAS_ROOT}/lib下需要包含openblas库
REFERENCE_CBLAS_ROOT,在${REFERENCE_CBLAS_ROOT}/include下需要包含cblas.h，在${REFERENCE_CBLAS_ROOT}/lib下需要包含cblas库
\ No newline at end of file
编译选项,描述,注意
MKL_ROOT,MKL的路径,${MKL_ROOT}/include下需要包含mkl.h，${MKL_ROOT}/lib目录下需要包含mkl_core，mkl_sequential和mkl_intel_lp64三个库。
ATLAS_ROOT,ATLAS的路径,${ATLAS_ROOT}/include下需要包含cblas.h，${ATLAS_ROOT}/lib下需要包含cblas和atlas两个库。
OPENBLAS_ROOT,OpenBLAS的路径,${OPENBLAS_ROOT}/include下需要包含cblas.h，${OPENBLAS_ROOT}/lib下需要包含openblas库。
REFERENCE_CBLAS_ROOT,REFERENCE BLAS的路径,${REFERENCE_CBLAS_ROOT}/include下需要包含cblas.h，${REFERENCE_CBLAS_ROOT}/lib下需要包含cblas库。
\ No newline at end of file
doc_cn/build_and_install/cmake/compile_options.csv
浏览文件 @ e4880016
选项,说明,默认值
WITH_GPU,是否编译GPU支持。,是否寻找到cuda工具链
WITH_DOUBLE,是否使用双精度浮点数。,否
WITH_DSO,是否使用运行时动态加载cuda动态库，而非静态加载cuda动态库。,是
WITH_AVX,是否编译含有AVX指令集的PaddlePaddle二进制,是
WITH_PYTHON,是否内嵌python解释器。可以方便嵌入式工作。,是
WITH_STYLE_CHECK,是否编译时进行代码风格检查,是
WITH_RDMA,是否开启RDMA支持,否
WITH_GLOG,是否使用GLOG，如果不使用则会使用一个简化版的日志实现。可以方便嵌入式工作。,取决于是否寻找到GLOG
WITH_GFLAGS,是否使用GFLAGS，如果不使用则会使用一个简化版的命令行参数解析。可以方便嵌入式工作。,取决于是否寻找到GFLAGS
WITH_TIMER,是否开启计时功能开启计时功能会导致运行略慢，打印的日志变多。但是方便调试和benchmark,否
WITH_TESTING,是否开启单元测试,取决于是否寻找到gtest
WITH_DOC,是否编译英文文档,否
WITH_DOC_CN,是否编译中文文档,否
WITH_SWIG_PY,是否编译python的swig接口，python的swig接口可以方便进行预测和定制化训练,取决于是否找到swig
选项,说明,默认值
WITH_GPU,是否支持GPU。,取决于是否寻找到CUDA工具链
WITH_DOUBLE,是否使用双精度浮点数。,否
WITH_DSO,是否运行时动态加载CUDA动态库，而非静态加载CUDA动态库。,是
WITH_AVX,是否编译含有AVX指令集的PaddlePaddle二进制文件,是
WITH_PYTHON,是否内嵌PYTHON解释器。方便今后的嵌入式移植工作。,是
WITH_STYLE_CHECK,是否编译时进行代码风格检查,是
WITH_RDMA,是否开启RDMA,否
WITH_GLOG,是否开启GLOG。如果不开启，则会使用一个简化版的日志，同时方便今后的嵌入式移植工作。,取决于是否寻找到GLOG
WITH_GFLAGS,是否使用GFLAGS。如果不开启，则会使用一个简化版的命令行参数解析器，同时方便今后的嵌入式移植工作。,取决于是否寻找到GFLAGS
WITH_TIMER,是否开启计时功能。如果开启会导致运行略慢，打印的日志变多，但是方便调试和测Benchmark,否
WITH_TESTING,是否开启单元测试,取决于是否寻找到GTEST
WITH_DOC,是否编译中英文文档,否
WITH_SWIG_PY,是否编译PYTHON的SWIG接口，该接口可用于预测和定制化训练,取决于是否寻找到SWIG
\ No newline at end of file
doc_cn/build_and_install/cmake/compile_options.rst
浏览文件 @ e4880016
设置PaddlePaddle的编译选项
==========================
PaddlePaddle的编译选项可以在调用cmake的时候设置。cmake是一个跨平台的编译脚本，调用
cmake可以将cmake项目文件，生成各个平台的makefile。详细的cmake使用方法可以参考
`cmake的官方文档 <https://cmake.org/cmake-tutorial>`_ 。
PaddlePaddle的编译选项是可以控制PaddlePaddle生成CPU/GPU版本二进制，链接何种blas等等。所有的
编译选项列表如下
PaddlePaddle的编译选项
----------------------
bool型的编译选项
++++++++++++++++
设置下列编译选项时，可以在cmake的命令行设置。使用 -D命令即可。例如
:code:`cmake -D WITH_GPU=OFF`
.. csv-table:: PaddlePaddle的bool型编译选项
:widths: 1, 7, 2
:file: compile_options.csv
blas相关的编译选项
++++++++++++++++++
PaddlePaddle可以使用 `MKL <https://software.intel.com/en-us/intel-mkl>`_ ，
`Atlas <http://math-atlas.sourceforge.net/>`_ ,
`OpenBlas <http://www.openblas.net/>`_ 和
`refference Blas <http://www.netlib.org/blas/>`_ ，任意一种cblas实现。
通过编译时指定路径来实现引用各种blas。
cmake编译时会首先在系统路径(/usr/lib\:/usr/local/lib)中寻找这些blas的实现。同时
也会读取相关路径变量来进行搜索。路径变量为\:
.. csv-table:: PaddlePaddle的cblas编译选项
:widths: 1, 9
:header: "编译选项", "描述"
:file: cblas_settings.csv
这些变量均可以使用 -D命令指定。例如 :code:`cmake -D MKL_ROOT=/opt/mkl/`。这些变
量也可以通过调用cmake命令前通过环境变量指定。例如
.. code-block:: bash
export MKL_ROOT=/opt/mkl
cmake
需要注意的是，这些变量只在第一次cmake的时候有效。如果在第一次cmake之后想要重新设
置这些变量，推荐清理( :code:`rm -rf` )掉编译目录后，再指定。
cuda/cudnn相关的编译选项
++++++++++++++++++++++++
PaddlePaddle可以使用 cudnn v2之后的任何一个cudnn版本来编译运行。但需要注意的是编译和
运行使用的cudnn尽量是同一个版本。推荐使用最新版本的cudnn v5.1。
在cmake配置时可以使用 :code:`CUDNN_ROOT` 来配置CUDNN的安装路径。使用的命令也是
-D，例如 :code:`cmake -D CUDNN_ROOT=/opt/cudnnv5` 。
需要注意的是，这些变量只在第一次cmake的时候有效。如果在第一次cmake之后想要重新设
置这些变量，推荐清理( :code:`rm -rf` )掉编译目录后，再指定。
PaddlePaddle的编译选项
======================
PaddlePaddle的编译选项，包括生成CPU/GPU二进制文件、链接何种BLAS库等。用户可在调用cmake的时候设置它们，详细的cmake使用方法可以参考 `官方文档 <https://cmake.org/cmake-tutorial>`_ 。
Bool型的编译选项
----------------
用户可在cmake的命令行中，通过使用 ``-D`` 命令设置该类编译选项，例如
.. code-block:: bash
cmake .. -DWITH_GPU=OFF
.. csv-table:: Bool型的编译选项
:widths: 1, 7, 2
:file: compile_options.csv
BLAS/CUDA/Cudnn的编译选项
--------------------------
BLAS
+++++
PaddlePaddle支持以下任意一种BLAS库：`MKL <https://software.intel.com/en-us/intel-mkl>`_ ，`ATLAS <http://math-atlas.sourceforge.net/>`_ ，`OpenBlAS <http://www.openblas.net/>`_ 和 `REFERENCE BLAS <http://www.netlib.org/blas/>`_ 。
.. csv-table:: BLAS路径相关的编译选项
:widths: 1, 2, 7
:file: cblas_settings.csv
CUDA/Cudnn
+++++++++++
PaddlePaddle可以使用cudnn v2之后的任何一个版本来编译运行，但尽量请保持编译和运行使用的cudnn是同一个版本。 我们推荐使用最新版本的cudnn v5.1。
编译选项的设置
++++++++++++++
PaddePaddle通过编译时指定路径来实现引用各种BLAS/CUDA/Cudnn库。cmake编译时，首先在系统路径(/usr/lib\:/usr/local/lib)中搜索这几个库，同时也会读取相关路径变量来进行搜索。 通过使用 ``-D`` 命令可以设置，例如
.. code-block:: bash
cmake .. -DMKL_ROOT=/opt/mkl/ -DCUDNN_ROOT=/opt/cudnnv5
注意：这几个编译选项的设置，只在第一次cmake的时候有效。如果之后想要重新设置，推荐清理整个编译目录（``rm -rf``）后，再指定。
\ No newline at end of file
paddle/api/Arguments.cpp
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "PaddleAPI.h"
#include "PaddleAPIPrivate.h"
......@@ -112,7 +111,7 @@ void Arguments::setSlotSequenceStartPositions(size_t idx,
}
void Arguments::setSlotSubSequenceStartPositions(
size_t idx, IVector *vec) throw(RangeError) {
size_t idx, IVector* vec) throw(RangeError) {
auto& a = m->getArg(idx);
auto& v = m->cast<paddle::IVector>(vec->getSharedPtr());
a.subSequenceStartPositions = std::make_shared<paddle::ICpuGpuVector>(v);
......
paddle/api/ConfigParser.cpp
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "PaddleAPI.h"
#include "PaddleAPIPrivate.h"
#include "paddle/trainer/Trainer.h"
......@@ -44,8 +43,7 @@ TrainerConfig* TrainerConfig::createFromTrainerConfigFile(
return retv;
}
TrainerConfig* TrainerConfig::createFromProtoString(
const std::string& str) {
TrainerConfig* TrainerConfig::createFromProtoString(const std::string& str) {
auto retv = new TrainerConfig();
paddle::TrainerConfig trainerConfigProto;
auto conf = std::make_shared<paddle::TrainerConfigHelper>(trainerConfigProto);
......
paddle/api/GradientMachine.cpp
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "PaddleAPI.h"
#include "PaddleAPIPrivate.h"
......@@ -27,7 +26,8 @@ GradientMachine::GradientMachine() : m(new GradientMachinePrivate()) {}
GradientMachine::~GradientMachine() { delete m; }
GradientMachine* GradientMachine::createFromPaddleModelPtr(
const void* confPtr, GradientMatchineCreateMode mode,
const void* confPtr,
GradientMatchineCreateMode mode,
const std::vector<int>& types) {
auto& conf = *(const paddle::ModelConfig*)(confPtr);
std::vector<ParameterType> realTypes;
......@@ -44,7 +44,8 @@ GradientMachine* GradientMachine::createFromPaddleModelPtr(
}
GradientMachine* GradientMachine::createByConfigProtoStr(
const std::string& protoStr, GradientMatchineCreateMode mode,
const std::string& protoStr,
GradientMatchineCreateMode mode,
const std::vector<int>& types) {
paddle::ModelConfig conf;
conf.ParseFromString(protoStr);
......@@ -56,13 +57,15 @@ GradientMachine* GradientMachine::createByConfigProtoStr(
}
GradientMachine* GradientMachine::createByModelConfig(
ModelConfig* conf, GradientMatchineCreateMode mode,
ModelConfig* conf,
GradientMatchineCreateMode mode,
const std::vector<int>& types) {
auto confPtr = &conf->m->conf->getModelConfig();
return GradientMachine::createFromPaddleModelPtr(confPtr, mode, types);
}
void GradientMachine::forward(const Arguments& inArgs, Arguments* outArgs,
void GradientMachine::forward(const Arguments& inArgs,
Arguments* outArgs,
PassType passType) {
auto& in =
m->cast<std::vector<paddle::Argument>>(inArgs.getInternalArgumentsPtr());
......@@ -99,7 +102,8 @@ void GradientMachine::backward(const UpdateCallback& callback) {
}
void GradientMachine::forwardBackward(const Arguments& inArgs,
Arguments* outArgs, PassType passType,
Arguments* outArgs,
PassType passType,
const UpdateCallback& callback) {
auto& in =
m->cast<std::vector<paddle::Argument>>(inArgs.getInternalArgumentsPtr());
......@@ -129,7 +133,7 @@ Parameter* GradientMachine::getParameter(size_t i) throw(RangeError) {
void GradientMachine::randParameters() { m->machine->randParameters(); }
Matrix* GradientMachine::getLayerOutput(const std::string& layerName) const
throw(UnsupportError) {
throw(UnsupportError) {
auto nn = std::dynamic_pointer_cast<paddle::NeuralNetwork>(m->machine);
if (nn) {
auto mat = nn->getLayerOutput(layerName);
......@@ -140,8 +144,11 @@ Matrix* GradientMachine::getLayerOutput(const std::string& layerName) const
}
SequenceGenerator* GradientMachine::asSequenceGenerator(
const std::vector<std::string>& dict, size_t begin_id, size_t end_id,
size_t max_length, size_t beam_size) {
const std::vector<std::string>& dict,
size_t begin_id,
size_t end_id,
size_t max_length,
size_t beam_size) {
SequenceGenerator* r =
SequenceGenerator::createByGradientMachineSharedPtr(&m->machine);
r->setDict(dict);
......
paddle/api/Internal.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#pragma once
#include "PaddleAPI.h"
......@@ -23,7 +22,8 @@ limitations under the License. */
template <typename T1, typename T2>
void staticCastVector(std::vector<T2>* dest, const std::vector<T1>& src) {
dest->resize(src.size());
std::transform(src.begin(), src.end(), dest->begin(), [](T1 t){
return static_cast<T2>(t);
});
std::transform(src.begin(),
src.end(),
dest->begin(),
[](T1 t) { return static_cast<T2>(t); });
}
paddle/api/Matrix.cpp
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "PaddleAPI.h"
#include "paddle/math/Matrix.h"
#include "paddle/math/SparseMatrix.h"
......@@ -44,17 +43,21 @@ Matrix* Matrix::createZero(size_t height, size_t width, bool useGpu) {
return m;
}
Matrix* Matrix::createDense(const std::vector<float>& data, size_t height,
size_t width, bool useGpu) {
Matrix* Matrix::createDense(const std::vector<float>& data,
size_t height,
size_t width,
bool useGpu) {
auto m = new Matrix();
m->m->mat = paddle::Matrix::create(height, width, useGpu);
m->m->mat->copyFrom(data.data(), data.size());
return m;
}
Matrix* Matrix::createDenseFromNumpy(float* data, int dim1, int dim2,
bool copy, bool useGpu)
throw (UnsupportError) {
Matrix* Matrix::createDenseFromNumpy(float* data,
int dim1,
int dim2,
bool copy,
bool useGpu) throw(UnsupportError) {
if (useGpu) {
/// Gpu mode only supports copy=True
if (!copy) {
......@@ -66,7 +69,9 @@ Matrix* Matrix::createDenseFromNumpy(float* data, int dim1, int dim2,
}
}
Matrix* Matrix::createCpuDenseFromNumpy(float* data, int dim1, int dim2,
Matrix* Matrix::createCpuDenseFromNumpy(float* data,
int dim1,
int dim2,
bool copy) {
auto m = new Matrix();
if (copy) {
......@@ -85,12 +90,20 @@ Matrix* Matrix::createGpuDenseFromNumpy(float* data, int dim1, int dim2) {
return m;
}
Matrix* Matrix::createSparse(size_t height, size_t width, size_t nnz,
bool isNonVal, bool isTrans, bool useGpu) {
Matrix* Matrix::createSparse(size_t height,
size_t width,
size_t nnz,
bool isNonVal,
bool isTrans,
bool useGpu) {
auto m = new Matrix();
m->m->mat = paddle::Matrix::createSparseMatrix(
height, width, nnz, isNonVal ? paddle::NO_VALUE : paddle::FLOAT_VALUE,
isTrans, useGpu);
height,
width,
nnz,
isNonVal ? paddle::NO_VALUE : paddle::FLOAT_VALUE,
isTrans,
useGpu);
return m;
}
......@@ -221,7 +234,8 @@ FloatArray Matrix::getData() const {
}
void Matrix::sparseCopyFrom(
const std::vector<int>& rows, const std::vector<int>& cols,
const std::vector<int>& rows,
const std::vector<int>& cols,
const std::vector<float>& vals) throw(UnsupportError) {
auto cpuSparseMat =
std::dynamic_pointer_cast<paddle::CpuSparseMatrix>(m->mat);
......@@ -240,7 +254,8 @@ void Matrix::sparseCopyFrom(
void* Matrix::getSharedPtr() const { return &m->mat; }
void Matrix::toNumpyMatInplace(float** view_data, int* dim1,
void Matrix::toNumpyMatInplace(float** view_data,
int* dim1,
int* dim2) throw(UnsupportError) {
auto cpuMat = std::dynamic_pointer_cast<paddle::CpuMatrix>(m->mat);
if (cpuMat) {
......@@ -251,7 +266,8 @@ void Matrix::toNumpyMatInplace(float** view_data, int* dim1,
throw UnsupportError();
}
}
void Matrix::copyToNumpyMat(float** view_m_data, int* dim1,
void Matrix::copyToNumpyMat(float** view_m_data,
int* dim1,
int* dim2) throw(UnsupportError) {
static_assert(sizeof(paddle::real) == sizeof(float),
"Currently PaddleAPI only support for single "
......@@ -269,8 +285,8 @@ void Matrix::copyToNumpyMat(float** view_m_data, int* dim1,
} else if (auto gpuMat = dynamic_cast<paddle::GpuMatrix*>(m->mat.get())) {
auto src = gpuMat->getData();
auto dest = *view_m_data;
hl_memcpy_device2host(dest, src,
sizeof(paddle::real) * (*dim1) * (*dim2));
hl_memcpy_device2host(
dest, src, sizeof(paddle::real) * (*dim1) * (*dim2));
} else {
LOG(WARNING) << "Unexpected Situation";
throw UnsupportError();
......@@ -278,7 +294,8 @@ void Matrix::copyToNumpyMat(float** view_m_data, int* dim1,
}
}
void Matrix::copyFromNumpyMat(float* data, int dim1,
void Matrix::copyFromNumpyMat(float* data,
int dim1,
int dim2) throw(UnsupportError, RangeError) {
if (isSparse()) {
throw UnsupportError();
......
paddle/api/PaddleAPI.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#pragma once
#include <stddef.h>
......@@ -61,8 +60,8 @@ class RangeError {};
/// Not support Error, such as access GPU memory directly, etc.
class UnsupportError : public std::runtime_error {
public:
UnsupportError() : std::runtime_error(" ") {};
UnsupportError(const std::string& message) : std::runtime_error(message) {};
UnsupportError() : std::runtime_error(" "){};
UnsupportError(const std::string& message) : std::runtime_error(message){};
};
/// This type will map to python's list of float.
......@@ -112,7 +111,8 @@ public:
/**
* Create A Matrix with height,width, which is filled by zero.
*/
static Matrix* createZero(size_t height, size_t width,
static Matrix* createZero(size_t height,
size_t width,
bool useGpu = isUsingGpu());
/**
......@@ -124,8 +124,11 @@ public:
*
* @note the default sparse type is SPARSE_CSR.
*/
static Matrix* createSparse(size_t height, size_t width, size_t nnz,
bool isNonVal = true, bool trans = false,
static Matrix* createSparse(size_t height,
size_t width,
size_t nnz,
bool isNonVal = true,
bool trans = false,
bool useGpu = isUsingGpu());
/**
......@@ -134,13 +137,17 @@ public:
* @param data list of float should be passed in python.
* @note the value will be copy into a new matrix.
*/
static Matrix* createDense(const std::vector<float>& data, size_t height,
size_t width, bool useGpu = isUsingGpu());
static Matrix* createDenseFromNumpy(float* data, int dim1, int dim2,
bool copy = true,
bool useGpu = isUsingGpu())
throw (UnsupportError);
static Matrix* createDense(const std::vector<float>& data,
size_t height,
size_t width,
bool useGpu = isUsingGpu());
static Matrix* createDenseFromNumpy(
float* data,
int dim1,
int dim2,
bool copy = true,
bool useGpu = isUsingGpu()) throw(UnsupportError);
/**
* Create Cpu Dense Matrix from numpy matrix, dtype=float32
......@@ -151,7 +158,9 @@ public:
* @param copy true if copy into a new matrix, false will create
* matrix inplace.
*/
static Matrix* createCpuDenseFromNumpy(float* data, int dim1, int dim2,
static Matrix* createCpuDenseFromNumpy(float* data,
int dim1,
int dim2,
bool copy = false);
/// Create Gpu Dense Matrix from numpy matrix, dtype=float32
......@@ -171,11 +180,13 @@ public:
* numpy_mat = m.toNumpyMat()
* @endcode
*/
void toNumpyMatInplace(float** view_data, int* dim1,
void toNumpyMatInplace(float** view_data,
int* dim1,
int* dim2) throw(UnsupportError);
/// Copy To numpy mat.
void copyToNumpyMat(float** view_m_data, int* dim1,
void copyToNumpyMat(float** view_m_data,
int* dim1,
int* dim2) throw(UnsupportError);
/// Copy From Numpy Mat
......@@ -248,15 +259,18 @@ public:
static Vector* create(const std::vector<float>& data,
bool useGpu = isUsingGpu());
static Vector* createVectorFromNumpy(float* data, int dim, bool copy = true,
bool useGpu = isUsingGpu())
throw (UnsupportError);
static Vector* createVectorFromNumpy(
float* data,
int dim,
bool copy = true,
bool useGpu = isUsingGpu()) throw(UnsupportError);
/**
* Create Cpu Vector from numpy array, which dtype=float32
*
* If copy is false, it will create vector inplace.
*/
static Vector* createCpuVectorFromNumpy(float* data, int dim,
static Vector* createCpuVectorFromNumpy(float* data,
int dim,
bool copy = false);
/// Create Gpu Vector from numpy array, which dtype=float32
......@@ -312,16 +326,19 @@ public:
static IVector* create(const std::vector<int>& data,
bool useGpu = isUsingGpu());
static IVector* createVectorFromNumpy(int* data, int dim, bool copy = true,
bool useGpu = isUsingGpu())
throw (UnsupportError);
static IVector* createVectorFromNumpy(
int* data,
int dim,
bool copy = true,
bool useGpu = isUsingGpu()) throw(UnsupportError);
/**
* Create Cpu IVector from numpy array, which dtype=int32
*
* If copy is false, it will create vector inplace
*/
static IVector* createCpuVectorFromNumpy(int* data, int dim,
static IVector* createCpuVectorFromNumpy(int* data,
int dim,
bool copy = false);
/**
* Create Gpu IVector from numpy array, which dtype=int32
......@@ -605,7 +622,8 @@ class ParameterTraverseCallback {
public:
~ParameterTraverseCallback();
void apply(const std::vector<Vector*>& vecs, const ParameterConfig& config,
void apply(const std::vector<Vector*>& vecs,
const ParameterConfig& config,
size_t sparseId);
private:
......@@ -638,7 +656,8 @@ public:
void finishBatch();
void update(const std::vector<Vector*>& vecs, const ParameterConfig& conf,
void update(const std::vector<Vector*>& vecs,
const ParameterConfig& conf,
size_t sparseId = NO_SPARSE_ID);
std::vector<int> getParameterTypes() const;
......@@ -678,7 +697,8 @@ public:
* model config by TrainerConfig
*/
static GradientMachine* createByModelConfig(
ModelConfig* conf, GradientMatchineCreateMode mode = CREATE_MODE_NORMAL,
ModelConfig* conf,
GradientMatchineCreateMode mode = CREATE_MODE_NORMAL,
const std::vector<int>& parameterTypes = defaultParamTypes);
/**
......@@ -701,7 +721,8 @@ public:
/**
* Combine forward/backward
*/
void forwardBackward(const Arguments& inArgs, Arguments* outArgs,
void forwardBackward(const Arguments& inArgs,
Arguments* outArgs,
PassType passType,
const UpdateCallback& callback = UpdateCallback());
......@@ -722,14 +743,17 @@ public:
*/
SequenceGenerator* asSequenceGenerator(
const std::vector<std::string>& dict = std::vector<std::string>(),
size_t begin_id = 0UL, size_t end_id = 0UL, size_t max_length = 100UL,
size_t begin_id = 0UL,
size_t end_id = 0UL,
size_t max_length = 100UL,
size_t beam_size = -1UL);
private:
GradientMachinePrivate* m;
static GradientMachine* createFromPaddleModelPtr(
const void* confPtr, GradientMatchineCreateMode mode,
const void* confPtr,
GradientMatchineCreateMode mode,
const std::vector<int>& types);
// Not to use c++ 11 init-list, so we use static var as function default arg.
......@@ -751,8 +775,8 @@ public:
/// Create A Trainer By TrainerConfig. using paddle command line.
static Trainer* createByCommandLine() throw(IOError);
static Trainer* create(TrainerConfig* optConfig, GradientMachine* gm)
throw(IOError);
static Trainer* create(TrainerConfig* optConfig,
GradientMachine* gm) throw(IOError);
/// Start training
void startTrain();
......
paddle/api/Parameter.cpp
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "PaddleAPI.h"
#include "paddle/parameter/Parameter.h"
......
paddle/api/ParameterOptimizer.cpp
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "PaddleAPI.h"
#include "PaddleAPIPrivate.h"
#include "paddle/parameter/ParameterOptimizer.h"
......@@ -32,17 +31,21 @@ struct ParameterTraverseCallbackPrivate {
const paddle::ParameterOptimizer::TraverseCallback& callback)
: callback(callback) {}
void apply(const std::vector<Vector*>& vecs, const ParameterConfig& conf,
void apply(const std::vector<Vector*>& vecs,
const ParameterConfig& conf,
size_t sparseId) {
std::vector<paddle::VectorPtr> real_vecs;
real_vecs.resize(vecs.size());
std::transform(vecs.begin(), vecs.end(), real_vecs.begin(), [](Vector* v) {
if (v) {
return *(paddle::VectorPtr*)(v->getSharedPtr());
} else {
return paddle::VectorPtr();
}
});
std::transform(vecs.begin(),
vecs.end(),
real_vecs.begin(),
[](Vector* v) {
if (v) {
return *(paddle::VectorPtr*)(v->getSharedPtr());
} else {
return paddle::VectorPtr();
}
});
paddle::ParameterConfig& real_conf =
*(paddle::ParameterConfig*)(const_cast<ParameterConfig&>(conf)
......@@ -86,10 +89,12 @@ void ParameterOptimizer::startBatch(size_t numSamplesProcessed) {
void ParameterOptimizer::finishBatch() { m->optimizer->finishBatch(); }
void ParameterOptimizer::update(const std::vector<Vector*>& vecs,
const ParameterConfig& conf, size_t sparseId) {
ParameterTraverseCallbackPrivate invoker([&](
const paddle::VectorPtr _vecs[], const paddle::ParameterConfig& config,
size_t sid = -1UL) { m->optimizer->update(_vecs, config, sid); });
const ParameterConfig& conf,
size_t sparseId) {
ParameterTraverseCallbackPrivate invoker(
[&](const paddle::VectorPtr _vecs[],
const paddle::ParameterConfig& config,
size_t sid = -1UL) { m->optimizer->update(_vecs, config, sid); });
invoker.apply(vecs, conf, sparseId);
}
......@@ -116,8 +121,9 @@ void ParameterTraverseCallback::apply(const std::vector<Vector*>& vecs,
ParameterTraverseCallback* ParameterOptimizer::needSpecialTraversal(
const ParameterConfig& config) const {
auto& param_config = *(paddle::ParameterConfig*)const_cast<ParameterConfig&>(
config).getRawPtr();
auto& param_config =
*(paddle::ParameterConfig*)const_cast<ParameterConfig&>(config)
.getRawPtr();
auto callback = m->optimizer->needSpecialTraversal(param_config);
if (callback) {
auto retCallback = new ParameterTraverseCallback();
......
paddle/api/SequenceGenerator.cpp
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "PaddleAPI.h"
#include "paddle/gserver/gradientmachines/GradientMachine.h"
#include "paddle/parameter/Argument.h"
......@@ -42,8 +41,10 @@ struct Path {
// position
static void findNBest(paddle::GradientMachine* gradMachine,
std::vector<paddle::Argument>& inArgs,
std::vector<Path>& finalPaths, size_t bos_id,
size_t eos_id, size_t max_length) {
std::vector<Path>& finalPaths,
size_t bos_id,
size_t eos_id,
size_t max_length) {
std::vector<Path> paths;
Path emptyPath;
paths.push_back(emptyPath);
......@@ -166,7 +167,8 @@ public:
if (id < getSize()) {
Path& p = (*path_)[id];
std::ostringstream sout;
std::transform(p.ids.begin(), p.ids.end(),
std::transform(p.ids.begin(),
p.ids.end(),
std::ostream_iterator<std::string>(sout, split ? " " : ""),
[&](int id) { return (*dict_)[id]; });
return sout.str();
......
paddle/api/Trainer.cpp
浏览文件 @ e4880016
......@@ -64,12 +64,11 @@ Trainer* Trainer::createByCommandLine() throw(IOError) {
Trainer::Trainer(TrainerConfig* config, GradientMachine* gm)
: m(new TrainerPrivate()) {
m->init(config->m->conf, /* testing= */false, gm ? gm->m->machine : nullptr);
m->init(config->m->conf, /* testing= */ false, gm ? gm->m->machine : nullptr);
}
Trainer* Trainer::create(TrainerConfig* config, GradientMachine* gm)
throw(IOError)
{
Trainer* Trainer::create(TrainerConfig* config,
GradientMachine* gm) throw(IOError) {
auto retv = new Trainer(config, gm);
if (retv->m->getConfig().IsInitialized()) {
return retv;
......@@ -134,15 +133,17 @@ void Trainer::finishTestPeriod() { m->finishTestPeriod(); }
Matrix* Trainer::getLayerOutput(const std::string& layerName) {
auto nn = std::dynamic_pointer_cast<paddle::NeuralNetwork>(
this->m->getGradientMachine());
this->m->getGradientMachine());
CHECK(nn) << "trainerInternal_.getGradientMachine() is not NeuralNetwork";
auto m = nn->getLayerOutput(layerName);
return Matrix::createByPaddleMatrixPtr(&m);
}
void Trainer::forwardOneBatch(size_t batchSize) { m->forwardOneBatch(batchSize); }
void Trainer::forwardOneBatch(size_t batchSize) {
m->forwardOneBatch(batchSize);
}
bool TrainerPrivate::forwardOneBatch(size_t batchSize) {
bool TrainerPrivate::forwardOneBatch(size_t batchSize) {
CHECK(dataProvider_) << "data_provider is not specified";
paddle::DataBatch dataBatch;
int num = dataProvider_->getNextBatch(batchSize, &dataBatch);
......@@ -156,7 +157,6 @@ bool TrainerPrivate::forwardOneBatch(size_t batchSize) {
void TrainerPrivate::forwardOneDataBatch(
const std::vector<paddle::Argument>& inArgs) {
std::vector<paddle::Argument>& outArgs = forwardOutput_;
if (config_->getOptConfig().use_sparse_remote_updater()) {
......
paddle/api/Util.cpp
浏览文件 @ e4880016
......@@ -37,13 +37,15 @@ FloatArray::FloatArray(const float* b, const size_t l)
IntArray::IntArray(const int* b, const size_t l, bool f)
: buf(b), length(l), needFree(f) {}
IntWithFloatArray::IntWithFloatArray(const float* v, const int* i, size_t l,
IntWithFloatArray::IntWithFloatArray(const float* v,
const int* i,
size_t l,
bool f)
: valBuf(v), idxBuf(i), length(l), needFree(f) {}
bool isUsingGpu() {return FLAGS_use_gpu;}
bool isUsingGpu() { return FLAGS_use_gpu; }
void setUseGpu(bool useGpu) {FLAGS_use_gpu = useGpu;}
void setUseGpu(bool useGpu) { FLAGS_use_gpu = useGpu; }
bool isGpuVersion() {
#ifdef PADDLE_ONLY_CPU
......
paddle/api/Vector.cpp
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include "PaddleAPI.h"
#include "paddle/math/Vector.h"
......@@ -39,8 +38,10 @@ IVector* IVector::create(const std::vector<int>& data, bool useGpu) {
return v;
}
IVector* IVector::createVectorFromNumpy(int* data, int dim, bool copy,
bool useGpu) throw (UnsupportError){
IVector* IVector::createVectorFromNumpy(int* data,
int dim,
bool copy,
bool useGpu) throw(UnsupportError) {
if (useGpu) {
/// if use gpu only copy=true is supported
if (!copy) {
......@@ -137,8 +138,8 @@ void IVector::copyToNumpyArray(int** view_m_data, int* dim1) {
if (auto cpuVec = dynamic_cast<paddle::CpuIVector*>(m->vec.get())) {
std::memcpy(*view_m_data, cpuVec->getData(), sizeof(int) * (*dim1));
} else if (auto gpuVec = dynamic_cast<paddle::GpuIVector*>(m->vec.get())) {
hl_memcpy_device2host(*view_m_data, gpuVec->getData(),
sizeof(int) * (*dim1));
hl_memcpy_device2host(
*view_m_data, gpuVec->getData(), sizeof(int) * (*dim1));
} else {
LOG(INFO) << "Unexpected situation";
}
......@@ -201,8 +202,10 @@ Vector* Vector::createByPaddleVectorPtr(void* ptr) {
}
}
Vector* Vector::createVectorFromNumpy(float* data, int dim, bool copy,
bool useGpu) throw (UnsupportError){
Vector* Vector::createVectorFromNumpy(float* data,
int dim,
bool copy,
bool useGpu) throw(UnsupportError) {
if (useGpu) {
/// if use gpu only copy=True is supported
if (!copy) {
......@@ -251,8 +254,8 @@ void Vector::copyToNumpyArray(float** view_m_data, int* dim1) {
if (auto cpuVec = dynamic_cast<paddle::CpuVector*>(m->vec.get())) {
std::memcpy(*view_m_data, cpuVec->getData(), sizeof(float) * (*dim1));
} else if (auto gpuVec = dynamic_cast<paddle::CpuVector*>(m->vec.get())) {
hl_memcpy_device2host(*view_m_data, gpuVec->getData(),
sizeof(float) * (*dim1));
hl_memcpy_device2host(
*view_m_data, gpuVec->getData(), sizeof(float) * (*dim1));
} else {
LOG(INFO) << "Unexpected situation";
}
......
paddle/cuda/CMakeLists.txt
浏览文件 @ e4880016
......@@ -81,5 +81,8 @@ else()
add_library(paddle_cuda ${CUDA_SOURCES})
endif()
add_style_check_target(paddle_cuda ${CUDA_SOURCES})
add_style_check_target(paddle_cuda ${CUDA_HEADERS})
add_style_check_target(paddle_cuda
${CUDA_SOURCES}
${CUDA_HEADERS}
${CUDA_DSO_SOURCES}
${CUDA_CXX_WITH_GPU_SOURCES})
paddle/cuda/include/hl_activation_functions.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_ACTIVATION_FUNCTIONS_H_
#define HL_ACTIVATION_FUNCTIONS_H_
......@@ -21,11 +20,8 @@ limitations under the License. */
/**
* Active functions: sigmoid, relu, tanh and linear.
*/
#define HPPL_ACTIVE_FUNCTION {hppl::sigmoid, \
hppl::relu, \
hppl::tanh, \
hppl::linear \
}
#define HPPL_ACTIVE_FUNCTION \
{ hppl::sigmoid, hppl::relu, hppl::tanh, hppl::linear }
namespace hppl {
......@@ -42,18 +38,18 @@ public:
#ifdef __NVCC__
namespace gpu {
static __device__ Active<real>::forward forward[] = HPPL_ACTIVE_FUNCTION;
static __device__ Active<real>::forward forward[] = HPPL_ACTIVE_FUNCTION;
static __device__ Active<real>::backward backward[] = HPPL_ACTIVE_FUNCTION;
}
#else
namespace cpu {
static Active<real>::forward forward[] = HPPL_ACTIVE_FUNCTION;
static Active<real>::forward forward[] = HPPL_ACTIVE_FUNCTION;
static Active<real>::backward backward[] = HPPL_ACTIVE_FUNCTION;
}
#ifdef __AVX__
namespace avx {
static Active<__m256>::forward forward[] = HPPL_ACTIVE_FUNCTION;
static Active<__m256>::forward forward[] = HPPL_ACTIVE_FUNCTION;
static Active<__m256>::backward backward[] = HPPL_ACTIVE_FUNCTION;
}
#endif
......
paddle/cuda/include/hl_aggregate.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_AGGREGATE_H_
#define HL_AGGREGATE_H_
......
paddle/cuda/include/hl_avx_functions.h
浏览文件 @ e4880016
......@@ -12,22 +12,21 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_AVX_FUNCTIONS_H_
#define HL_AVX_FUNCTIONS_H_
#include <immintrin.h>
namespace hppl {
__m256 relu(const __m256 a);
__m256 sigmoid(const __m256 a);
__m256 tanh(const __m256 a);
__m256 linear(const __m256 a);
__m256 relu(const __m256 a, const __m256 b);
__m256 sigmoid(const __m256 a, const __m256 b);
__m256 tanh(const __m256 a, const __m256 b);
__m256 linear(const __m256 a, const __m256 b);
__m256 relu(const __m256 a);
__m256 sigmoid(const __m256 a);
__m256 tanh(const __m256 a);
__m256 linear(const __m256 a);
__m256 relu(const __m256 a, const __m256 b);
__m256 sigmoid(const __m256 a, const __m256 b);
__m256 tanh(const __m256 a, const __m256 b);
__m256 linear(const __m256 a, const __m256 b);
} // namespace hppl
#endif // HL_AVX_FUNCTIONS_H_
paddle/cuda/include/hl_base.h
浏览文件 @ e4880016
......@@ -12,8 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_BASE_H_
#define HL_BASE_H_
......@@ -33,36 +31,36 @@ limitations under the License. */
* HPPL_STREAM_DEFAULT is HPPL default stream.
*/
typedef enum {
HPPL_STREAM_DEFAULT = 0, /* Thread Default Stream*/
HPPL_STREAM_1 = 1,
HPPL_STREAM_2 = 2,
HPPL_STREAM_3 = 3,
HPPL_STREAM_4 = 4,
HPPL_THREAD_STREAM_1 = 5,
HPPL_THREAD_STREAM_2 = 6,
HPPL_THREAD_STREAM_3 = 7,
HPPL_THREAD_STREAM_4 = 8,
HPPL_STREAM_END
HPPL_STREAM_DEFAULT = 0, /* Thread Default Stream*/
HPPL_STREAM_1 = 1,
HPPL_STREAM_2 = 2,
HPPL_STREAM_3 = 3,
HPPL_STREAM_4 = 4,
HPPL_THREAD_STREAM_1 = 5,
HPPL_THREAD_STREAM_2 = 6,
HPPL_THREAD_STREAM_3 = 7,
HPPL_THREAD_STREAM_4 = 8,
HPPL_STREAM_END
} hl_stream_t;
/**
* @brief HPPL activation mode.
*/
typedef enum {
HL_ACTIVATION_SIGMOID = 0,
HL_ACTIVATION_RELU = 1,
HL_ACTIVATION_TANH = 2,
HL_ACTIVATION_LINEAR = 3,
HL_ACTIVATION_END
HL_ACTIVATION_SIGMOID = 0,
HL_ACTIVATION_RELU = 1,
HL_ACTIVATION_TANH = 2,
HL_ACTIVATION_LINEAR = 3,
HL_ACTIVATION_END
} hl_activation_mode_t;
/**
* @brief Transpose type.
*/
typedef enum {
HPPL_OP_N = 0, /* transpose */
HPPL_OP_T = 1, /* non transpose */
HPPL_OP_END
HPPL_OP_N = 0, /* transpose */
HPPL_OP_T = 1, /* non transpose */
HPPL_OP_END
} hl_trans_op_t;
/**
......@@ -148,23 +146,21 @@ typedef struct {
* @brief Sparse matrix value type.
*/
typedef enum {
HL_NO_VALUE = 0, /* matrix values only 0 or 1 */
HL_FLOAT_VALUE = 1,
HL_VALUE_END
HL_NO_VALUE = 0, /* matrix values only 0 or 1 */
HL_FLOAT_VALUE = 1,
HL_VALUE_END
} hl_matrix_value_t;
/**
* @brief HPPL matrix format.
*/
typedef enum {
HL_SPARSE_CSR = 0,
HL_SPARSE_CSC = 1,
HL_SPARSE_END
HL_SPARSE_CSR = 0,
HL_SPARSE_CSC = 1,
HL_SPARSE_END
} hl_matrix_format_t;
typedef struct _hl_matrix_s * hl_matrix_s;
typedef struct _hl_matrix_s *hl_matrix_s;
/**
* @brief HPPL sparse matrix.
......@@ -177,12 +173,12 @@ typedef struct _hl_matrix_s * hl_matrix_s;
* @param nnz nonzero values of sparse matrix.
*/
typedef struct {
hl_matrix_s matrix;
hl_matrix_format_t format;
hl_matrix_value_t type;
int rows;
int cols;
size_t nnz;
hl_matrix_s matrix;
hl_matrix_format_t format;
hl_matrix_value_t type;
int rows;
int cols;
size_t nnz;
} _hl_sparse_matrix_s, *hl_sparse_matrix_s;
#ifndef PADDLE_TYPE_DOUBLE
......@@ -195,7 +191,7 @@ typedef struct {
*
* HL_FLOAT_MIN: 1.17549435e-38F
*/
#define HL_FLOAT_MAX 3.40282347e+38F
#define HL_FLOAT_MAX 3.40282347e+38F
/**
* if real == double
*
......@@ -203,20 +199,18 @@ typedef struct {
*
* HL_FLOAT_MIN: 2.2250738585072014e-308
*/
#define HL_FLOAT_MIN 1.17549435e-38F
#define HL_FLOAT_MIN 1.17549435e-38F
#else
#define HL_FLOAT_MAX 1.7976931348623157e+308
#define HL_FLOAT_MIN 2.2250738585072014e-308
#define HL_FLOAT_MAX 1.7976931348623157e+308
#define HL_FLOAT_MIN 2.2250738585072014e-308
#endif
/**
* The maximum input value for exp, used to avoid overflow problem.
*
* Currently only used for tanh function.
*/
#define EXP_MAX_INPUT 40.0
#define EXP_MAX_INPUT 40.0
/**
* @brief DIVUP(x, y) is similar to ceil(x / y).
......@@ -224,7 +218,7 @@ typedef struct {
* the size of blockDim.
*/
#ifndef DIVUP
#define DIVUP(x, y) (((x) + (y) - 1) / (y))
#define DIVUP(x, y) (((x) + (y)-1) / (y))
#endif
#ifdef __NVCC__
......@@ -233,7 +227,7 @@ typedef struct {
#include "hl_cuda.h"
#include "cuda_runtime.h"
extern __thread bool g_sync_flag;
extern __thread bool g_sync_flag;
extern __thread cudaStream_t default_stream;
#define STREAM_DEFAULT default_stream
......@@ -241,16 +235,15 @@ extern __thread cudaStream_t default_stream;
* @brief Check cuda kernel execution.
* @param msg error string
*/
#define CHECK_SYNC(msg) \
if (true == g_sync_flag) { \
hl_stream_synchronize(HPPL_STREAM_DEFAULT); \
cudaError_t err \
= (cudaError_t)hl_get_device_last_error(); \
CHECK_EQ(cudaSuccess, err) << "[" << msg << "] " \
<< "CUDA error: " \
<< hl_get_device_error_string((size_t)err); \
#define CHECK_SYNC(msg) \
if (true == g_sync_flag) { \
hl_stream_synchronize(HPPL_STREAM_DEFAULT); \
cudaError_t err = (cudaError_t)hl_get_device_last_error(); \
CHECK_EQ(cudaSuccess, err) \
<< "[" << msg << "] " \
<< "CUDA error: " << hl_get_device_error_string((size_t)err); \
}
#endif /* __NVCC__ */
#endif /* __NVCC__ */
#endif /* HL_BASE_H_ */
#endif /* HL_BASE_H_ */
paddle/cuda/include/hl_batch_transpose.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_BATCH_TRANSPOSE_H_
#define HL_BATCH_TRANSPOSE_H_
......@@ -31,10 +30,7 @@ limitations under the License. */
* order. Each batch has height * width data, which are
* arranged in height-first (or row-first) manner.
*/
extern void batchTranspose(const real* input,
real* output,
int width,
int height,
int batchSize);
extern void batchTranspose(
const real* input, real* output, int width, int height, int batchSize);
#endif // HL_BATCH_TRANSPOSE_H_
paddle/cuda/include/hl_cnn.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_CNN_H_
#define HL_CNN_H_
......@@ -37,15 +36,21 @@ limitations under the License. */
* @param[in] alpha
* @param[in] beta
*/
extern void hl_shrink_col2feature(
const real * dataCol, size_t channels,
size_t height, size_t width,
size_t blockH, size_t blockW,
size_t strideH, size_t strideW,
size_t paddingH, size_t paddingW,
size_t outputH, size_t outputW,
real* dataIm,
real alpha = 1.0f, real beta = 0.0f);
extern void hl_shrink_col2feature(const real* dataCol,
size_t channels,
size_t height,
size_t width,
size_t blockH,
size_t blockW,
size_t strideH,
size_t strideW,
size_t paddingH,
size_t paddingW,
size_t outputH,
size_t outputW,
real* dataIm,
real alpha = 1.0f,
real beta = 0.0f);
/**
* @brief Expand feature to column.
......@@ -65,14 +70,19 @@ extern void hl_shrink_col2feature(
* @param[out] dataCol expand data.
*
*/
extern void hl_expand_feature2col(
const real* dataIm, size_t channels,
size_t height, size_t width,
size_t blockH, size_t blockW,
size_t strideH, size_t strideW,
size_t paddingH, size_t paddingW,
size_t outputH, size_t outputW,
real* dataCol);
extern void hl_expand_feature2col(const real* dataIm,
size_t channels,
size_t height,
size_t width,
size_t blockH,
size_t blockW,
size_t strideH,
size_t strideW,
size_t paddingH,
size_t paddingW,
size_t outputH,
size_t outputW,
real* dataCol);
/**
* @brief Maximum pool forward.
......@@ -94,15 +104,21 @@ extern void hl_expand_feature2col(
* @param[in] tgtStride stride between output data samples.
*
*/
extern void hl_maxpool_forward(
const int frameCnt, const real* inputData,
const int channels,
const int height, const int width,
const int pooledH, const int pooledW,
const int sizeX, const int sizeY,
const int strideH, const int strideW,
const int paddingH, const int paddingW,
real* tgtData, const int tgtStride);
extern void hl_maxpool_forward(const int frameCnt,
const real* inputData,
const int channels,
const int height,
const int width,
const int pooledH,
const int pooledW,
const int sizeX,
const int sizeY,
const int strideH,
const int strideW,
const int paddingH,
const int paddingW,
real* tgtData,
const int tgtStride);
/**
* @brief Maximum pool backward.
......@@ -125,20 +141,28 @@ extern void hl_maxpool_forward(
* @param[in] paddingH padding height.
* @param[in] paddingW padding width.
* @param[out] targetGrad output grad.
* @param[in] outStride stride between output data samples.
* @param[in] outStride stride between output data samples.
*
*/
extern void hl_maxpool_backward(
const int frameCnt, const real* inputData,
const real* outData, const real* outGrad,
const int channels, const int height,
const int width,
const int pooledH, const int pooledW,
const int sizeX, const int sizeY,
const int strideH, const int strideW,
const int paddingH, const int paddingW,
real scaleA, real scaleB,
real* targetGrad, const int outStride);
extern void hl_maxpool_backward(const int frameCnt,
const real* inputData,
const real* outData,
const real* outGrad,
const int channels,
const int height,
const int width,
const int pooledH,
const int pooledW,
const int sizeX,
const int sizeY,
const int strideH,
const int strideW,
const int paddingH,
const int paddingW,
real scaleA,
real scaleB,
real* targetGrad,
const int outStride);
/**
* @brief Averge pool forward.
......@@ -160,15 +184,21 @@ extern void hl_maxpool_backward(
* @param[in] tgtStride stride between output data samples.
*
*/
extern void hl_avgpool_forward(
const int frameCnt, const real* inputData,
const int channels,
const int height, const int width,
const int pooledH, const int pooledW,
const int sizeX, const int sizeY,
const int strideH, const int strideW,
const int paddingH, const int paddingW,
real* tgtData, const int tgtStride);
extern void hl_avgpool_forward(const int frameCnt,
const real* inputData,
const int channels,
const int height,
const int width,
const int pooledH,
const int pooledW,
const int sizeX,
const int sizeY,
const int strideH,
const int strideW,
const int paddingH,
const int paddingW,
real* tgtData,
const int tgtStride);
/**
* @brief Maximum pool backward.
......@@ -189,19 +219,26 @@ extern void hl_avgpool_forward(
* @param[in] scaleA scale.
* @param[in] scaleB scale.
* @param[out] backGrad output grad.
* @param[in] outStride stride between output data samples.
* @param[in] outStride stride between output data samples.
*
*/
extern void hl_avgpool_backward(
const int frameCnt, const real* outGrad,
const int channels, const int height,
const int width,
const int pooledH, const int pooledW,
const int sizeX, const int sizeY,
const int strideH, const int strideW,
int paddingH, int paddingW,
real scaleA, real scaleB,
real* backGrad, const int outStride);
extern void hl_avgpool_backward(const int frameCnt,
const real* outGrad,
const int channels,
const int height,
const int width,
const int pooledH,
const int pooledW,
const int sizeX,
const int sizeY,
const int strideH,
const int strideW,
int paddingH,
int paddingW,
real scaleA,
real scaleB,
real* backGrad,
const int outStride);
/**
* @brief Cross-map-respose normalize forward.
......@@ -218,10 +255,16 @@ extern void hl_avgpool_backward(
* @param[in] beta scale.
*
*/
extern void hl_CMRNorm_forward(
size_t frameCnt, const real* in, real* scale, real* out,
size_t channels, size_t height, size_t width, size_t sizeX,
real alpha, real beta);
extern void hl_CMRNorm_forward(size_t frameCnt,
const real* in,
real* scale,
real* out,
size_t channels,
size_t height,
size_t width,
size_t sizeX,
real alpha,
real beta);
/**
* @brief Cross-map-respose normalize backward.
......@@ -240,11 +283,18 @@ extern void hl_CMRNorm_forward(
* @param[in] beta scale.
*
*/
extern void hl_CMRNorm_backward(
size_t frameCnt, const real* inV, const real* scale,
const real* outV, const real* outDiff, real *inDiff,
size_t channels, size_t height, size_t width, size_t sizeX,
real alpha, real beta);
extern void hl_CMRNorm_backward(size_t frameCnt,
const real* inV,
const real* scale,
const real* outV,
const real* outDiff,
real* inDiff,
size_t channels,
size_t height,
size_t width,
size_t sizeX,
real alpha,
real beta);
/**
* @brief Bilinear interpolation forward.
......@@ -278,24 +328,24 @@ extern void hl_bilinear_forward(const real* inData,
const real ratioH,
const real ratioW);
/**
* @brief Bilinear interpolation backward.
*
* @param[out] inGrad input gradient.
* @param[in] inImgH input image height.
* @param[in] inImgW input image width.
* @param[in] inputH input batchSize.
* @param[in] inputW input image data dim.
* @param[in] outGrad output gradient.
* @param[in] outImgH output image height.
* @param[in] outImgW output image width.
* @param[in] outputH output batchSize.
* @param[in] outputW output image data dim.
* @param[in] numChannels number of channels.
* @param[in] ratioH inImgH / outImgH.
* @param[in] ratioW inImgW / outImgW.
*
*/
/**
* @brief Bilinear interpolation backward.
*
* @param[out] inGrad input gradient.
* @param[in] inImgH input image height.
* @param[in] inImgW input image width.
* @param[in] inputH input batchSize.
* @param[in] inputW input image data dim.
* @param[in] outGrad output gradient.
* @param[in] outImgH output image height.
* @param[in] outImgW output image width.
* @param[in] outputH output batchSize.
* @param[in] outputW output image data dim.
* @param[in] numChannels number of channels.
* @param[in] ratioH inImgH / outImgH.
* @param[in] ratioW inImgW / outImgW.
*
*/
extern void hl_bilinear_backward(real* inGrad,
const size_t inImgH,
const size_t inImgW,
......@@ -321,9 +371,13 @@ extern void hl_bilinear_backward(real* inGrad,
* @param[in] featLen feature length = image height * image width.
* @param[in] groups number of groups.
*/
extern void hl_maxout_forward(
const real* inData, real* outData, int* idData,
size_t batchSize, size_t size, size_t featLen, size_t groups);
extern void hl_maxout_forward(const real* inData,
real* outData,
int* idData,
size_t batchSize,
size_t size,
size_t featLen,
size_t groups);
/**
* @brief MaxOut backward.
......@@ -336,8 +390,12 @@ extern void hl_maxout_forward(
* @param[in] featLen feature length = image height * image width.
* @param[in] groups number of groups.
*/
extern void hl_maxout_backward(
real* inGrad, const real* outGrad, const int* idData,
size_t batchSize, size_t size, size_t featLen, size_t groups);
extern void hl_maxout_backward(real* inGrad,
const real* outGrad,
const int* idData,
size_t batchSize,
size_t size,
size_t featLen,
size_t groups);
#endif /* HL_CNN_H_ */
paddle/cuda/include/hl_cuda.h
浏览文件 @ e4880016
......@@ -12,18 +12,16 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_CUDA_H_
#define HL_CUDA_H_
#include "hl_base.h"
#include <string>
#include "hl_base.h"
/**
* @brief HPPL event.
*/
typedef struct _hl_event_st * hl_event_t;
typedef struct _hl_event_st *hl_event_t;
/**
* @brief return cuda runtime api version.
......@@ -42,7 +40,7 @@ extern void hl_start();
* if device is NULL, will start all GPU.
* @param[in] number number of devices.
*/
extern void hl_specify_devices_start(int* device, int number);
extern void hl_specify_devices_start(int *device, int number);
/**
* @brief Queries if a device may directly access a peer device's memory.
......@@ -126,7 +124,7 @@ extern int hl_get_device();
*
* @return dest_d pointer to device memory.
*/
extern void* hl_malloc_device(size_t size);
extern void *hl_malloc_device(size_t size);
/**
* @brief Free device memory.
......@@ -143,7 +141,7 @@ extern void hl_free_mem_device(void *dest_d);
*
* @return dest_h pointer to host memory.
*/
extern void* hl_malloc_host(size_t size);
extern void *hl_malloc_host(size_t size);
/**
* @brief Free host page-lock memory.
......@@ -228,9 +226,9 @@ extern void hl_srand(unsigned int seed);
* @param[in] stream stream id.
*/
extern void hl_memcpy_async(void *dst,
void *src,
size_t size,
hl_stream_t stream);
void *src,
size_t size,
hl_stream_t stream);
/**
* @brief Waits for stream tasks to complete.
......@@ -261,8 +259,7 @@ extern void hl_destroy_event(hl_event_t event);
*
* @return time Time between start and end in ms.
*/
extern float hl_event_elapsed_time(hl_event_t start,
hl_event_t end);
extern float hl_event_elapsed_time(hl_event_t start, hl_event_t end);
/**
* @brief Records an event.
......@@ -300,7 +297,7 @@ extern void hl_set_device_flags_block();
/**
* @brief Returns the last error string from a cuda runtime call.
*/
extern const char* hl_get_device_error_string();
extern const char *hl_get_device_error_string();
/**
* @brief Returns the last error string from a cuda runtime call.
......@@ -309,7 +306,7 @@ extern const char* hl_get_device_error_string();
*
* @see hl_get_device_last_error()
*/
extern const char* hl_get_device_error_string(size_t err);
extern const char *hl_get_device_error_string(size_t err);
/**
* @brief Returns the last error number.
......
paddle/cuda/include/hl_cuda_cublas.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_CUDA_CUBLAS_H_
#define HL_CUDA_CUBLAS_H_
......@@ -29,12 +28,8 @@ limitations under the License. */
* @param[in] ldc the first dimension of C_d.
*
*/
extern void hl_matrix_transpose(real *A_d,
real *C_d,
int dimM,
int dimN,
int lda,
int ldc);
extern void hl_matrix_transpose(
real *A_d, real *C_d, int dimM, int dimN, int lda, int ldc);
/*
* @brief Matrix transpose, while lda = dimN, ldc = dimM.
......@@ -45,10 +40,7 @@ extern void hl_matrix_transpose(real *A_d,
* @param[in] dimN matrix width.
*
*/
extern void hl_matrix_transpose(real *A_d,
real *C_d,
int dimM,
int dimN);
extern void hl_matrix_transpose(real *A_d, real *C_d, int dimM, int dimN);
/*
* @brief Matrix inverse
......@@ -60,11 +52,7 @@ extern void hl_matrix_transpose(real *A_d,
* @param[in] ldc the first dimension of C_d
*
*/
extern void hl_matrix_inverse(real *A_d,
real *C_d,
int dimN,
int lda,
int ldc);
extern void hl_matrix_inverse(real *A_d, real *C_d, int dimN, int lda, int ldc);
/**
* @brief C_d = alpha*(op(A_d) * op(B_d)) + beta*C_d
......@@ -84,12 +72,19 @@ extern void hl_matrix_inverse(real *A_d,
* @param[in] ldc the first dimension of C_d.
*
*/
extern void hl_matrix_mul(real *A_d, hl_trans_op_t transa,
real *B_d, hl_trans_op_t transb,
extern void hl_matrix_mul(real *A_d,
hl_trans_op_t transa,
real *B_d,
hl_trans_op_t transb,
real *C_d,
int dimM, int dimN, int dimK,
real alpha, real beta,
int lda, int ldb, int ldc);
int dimM,
int dimN,
int dimK,
real alpha,
real beta,
int lda,
int ldb,
int ldc);
/**
* @brief C_d = alpha*(op(A_d) * op(B_d)) + beta*C_d
......@@ -106,11 +101,16 @@ extern void hl_matrix_mul(real *A_d, hl_trans_op_t transa,
* @param[in] beta scalar used for multiplication.
*
*/
extern void hl_matrix_mul(real *A_d, hl_trans_op_t transa,
real *B_d, hl_trans_op_t transb,
extern void hl_matrix_mul(real *A_d,
hl_trans_op_t transa,
real *B_d,
hl_trans_op_t transb,
real *C_d,
int dimM, int dimN, int dimK,
real alpha, real beta);
int dimM,
int dimN,
int dimK,
real alpha,
real beta);
/**
* @brief This function performs the matrix-vector multiplication.
......@@ -132,11 +132,17 @@ extern void hl_matrix_mul(real *A_d, hl_trans_op_t transa,
*
*/
extern void hl_matrix_mul_vector(real *A_d, hl_trans_op_t trans,
real *B_d, real *C_d,
int dimM, int dimN,
real alpha, real beta,
int lda, int incb, int incc);
extern void hl_matrix_mul_vector(real *A_d,
hl_trans_op_t trans,
real *B_d,
real *C_d,
int dimM,
int dimN,
real alpha,
real beta,
int lda,
int incb,
int incc);
/**
* @brief This function performs the matrix-vector multiplication.
......@@ -154,9 +160,13 @@ extern void hl_matrix_mul_vector(real *A_d, hl_trans_op_t trans,
* @param[in] beta scalar used for multiplication.
*
*/
extern void hl_matrix_mul_vector(real *A_d, hl_trans_op_t trans,
real *B_d, real *C_d,
int dimM, int dimN,
real alpha, real beta);
extern void hl_matrix_mul_vector(real *A_d,
hl_trans_op_t trans,
real *B_d,
real *C_d,
int dimM,
int dimN,
real alpha,
real beta);
#endif /* HL_CUDA_CUBLAS_H_ */
paddle/cuda/include/hl_cuda_cudnn.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_CUDA_CUDNN_H_
#define HL_CUDA_CUDNN_H_
......@@ -22,7 +21,7 @@ limitations under the License. */
* hppl pooling mode
*/
typedef enum {
HL_POOLING_MAX = 0,
HL_POOLING_MAX = 0,
// average includes padded values
HL_POOLING_AVERAGE = 1,
// average does not include padded values
......@@ -324,17 +323,16 @@ extern void hl_convolution_forward_add_bias(hl_tensor_descriptor bias,
* @param[in] sizeInBytes gpu workspace size (bytes).
* @param[in] convBwdFilterAlgo backward filter algorithm.
*/
extern void hl_convolution_backward_filter(
hl_tensor_descriptor input,
real* input_data,
hl_tensor_descriptor output,
real* output_grad_data,
hl_filter_descriptor filter,
real* filter_grad_data,
hl_convolution_descriptor conv,
void* gpuWorkSpace,
size_t sizeInBytes,
int convBwdFilterAlgo);
extern void hl_convolution_backward_filter(hl_tensor_descriptor input,
real* input_data,
hl_tensor_descriptor output,
real* output_grad_data,
hl_filter_descriptor filter,
real* filter_grad_data,
hl_convolution_descriptor conv,
void* gpuWorkSpace,
size_t sizeInBytes,
int convBwdFilterAlgo);
/**
* @brief convolution backward data(calculate input image grad data).
......@@ -350,17 +348,16 @@ extern void hl_convolution_backward_filter(
* @param[in] sizeInBytes gpu workspace size (bytes).
* @param[in] convBwdDataAlgo backward data algorithm.
*/
extern void hl_convolution_backward_data(
hl_tensor_descriptor input,
real* input_data_grad,
hl_tensor_descriptor output,
real* output_grad_data,
hl_filter_descriptor filter,
real* filter_data,
hl_convolution_descriptor conv,
void* gpuWorkSpace,
size_t sizeInBytes,
int convBwdDataAlgo);
extern void hl_convolution_backward_data(hl_tensor_descriptor input,
real* input_data_grad,
hl_tensor_descriptor output,
real* output_grad_data,
hl_filter_descriptor filter,
real* filter_data,
hl_convolution_descriptor conv,
void* gpuWorkSpace,
size_t sizeInBytes,
int convBwdDataAlgo);
/**
* @brief convolution backward bias(calculate bias grad data).
......@@ -383,8 +380,8 @@ extern void hl_convolution_backward_bias(hl_tensor_descriptor bias,
* @param[in] height matrix height.
* @param[in] width matrix width.
*/
extern void hl_softmax_forward(real *input,
real *output,
extern void hl_softmax_forward(real* input,
real* output,
int height,
int width);
......@@ -396,8 +393,8 @@ extern void hl_softmax_forward(real *input,
* @param[in] height matrix height.
* @param[in] width matrix width.
*/
extern void hl_softmax_backward(real *output_value,
real *output_grad,
extern void hl_softmax_backward(real* output_value,
real* output_grad,
int height,
int width);
......@@ -426,18 +423,18 @@ extern void hl_softmax_backward(real *output_value,
*
*/
extern void hl_batch_norm_forward_training(hl_tensor_descriptor inputDesc,
real *input,
real* input,
hl_tensor_descriptor outputDesc,
real *output,
real* output,
hl_tensor_descriptor bnParamDesc,
real *scale,
real *bias,
real* scale,
real* bias,
double factor,
real *runningMean,
real *runningInvVar,
real* runningMean,
real* runningInvVar,
double epsilon,
real *savedMean,
real *savedVar);
real* savedMean,
real* savedVar);
/**
* @brief cudnn batch norm forward.
......@@ -463,14 +460,14 @@ extern void hl_batch_norm_forward_training(hl_tensor_descriptor inputDesc,
*
*/
extern void hl_batch_norm_forward_inference(hl_tensor_descriptor inputDesc,
real *input,
real* input,
hl_tensor_descriptor outputDesc,
real *output,
real* output,
hl_tensor_descriptor bnParamDesc,
real *scale,
real *bias,
real *estimatedMean,
real *estimatedVar,
real* scale,
real* bias,
real* estimatedMean,
real* estimatedVar,
double epsilon);
/**
......@@ -483,7 +480,8 @@ extern void hl_batch_norm_forward_inference(hl_tensor_descriptor inputDesc,
* @param[in] inGradDesc input tensor descriptor desc.
* @param[in] inGrad input data.
* @param[in] dBnParamDesc tensor descriptor desc.
* bnScale, bnBias, running mean/var, save_mean/var.
* bnScale, bnBias, running mean/var,
* save_mean/var.
* @param[in] scale batch normalization scale parameter (in original
* paper scale is referred to as gamma).
* @param[in] scaleGrad batch normalization scale parameter (in original
......@@ -497,17 +495,17 @@ extern void hl_batch_norm_forward_inference(hl_tensor_descriptor inputDesc,
*
*/
extern void hl_batch_norm_backward(hl_tensor_descriptor inputDesc,
real *input,
real* input,
hl_tensor_descriptor outGradDesc,
real *outGrad,
real* outGrad,
hl_tensor_descriptor inGradDesc,
real *inGrad,
real* inGrad,
hl_tensor_descriptor dBnParamDesc,
real *scale,
real *scaleGrad,
real *biasGrad,
real* scale,
real* scaleGrad,
real* biasGrad,
double epsilon,
real *savedMean,
real *savedInvVar);
real* savedMean,
real* savedInvVar);
#endif // HL_CUDA_CUDNN_H_
paddle/cuda/include/hl_dso_loader.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_DSO_LOADER_H_
#define HL_DSO_LOADER_H_
......
paddle/cuda/include/hl_functions.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_FUNCTIONS_H_
#define HL_FUNCTIONS_H_
......@@ -21,30 +20,30 @@ limitations under the License. */
/**
* sigmoid threshold maximum
*/
#define SIGMOID_THRESHOLD_MIN -40.0
#define SIGMOID_THRESHOLD_MIN -40.0
/**
* sigmoid threshold minimum
*/
#define SIGMOID_THRESHOLD_MAX 13.0
#define SIGMOID_THRESHOLD_MAX 13.0
#ifndef __NVCC__
namespace hppl {
/*
* forward activation
*/
real relu(const real a);
real sigmoid(const real a);
real tanh(const real a);
real linear(const real a);
/*
* backward activation
*/
real relu(const real a, const real b);
real sigmoid(const real a, const real b);
real tanh(const real a, const real b);
real linear(const real a, const real b);
/*
* forward activation
*/
real relu(const real a);
real sigmoid(const real a);
real tanh(const real a);
real linear(const real a);
/*
* backward activation
*/
real relu(const real a, const real b);
real sigmoid(const real a, const real b);
real tanh(const real a, const real b);
real linear(const real a, const real b);
} // namespace hppl
#ifdef __AVX__
......
paddle/cuda/include/hl_gpu.h
浏览文件 @ e4880016
......@@ -12,7 +12,6 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#ifndef HL_GPU_H_
#define HL_GPU_H_
......
paddle/cuda/include/hl_lstm.h
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/include/hl_matrix.h
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/include/hl_sequence.h
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/include/hl_sparse.h
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/include/hl_time.h
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/include/hl_top_k.h
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/avx_mathfun.h
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/hl_avx_functions.cc
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/hl_cpu_functions.cc
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/hl_cuda_cublas.cc
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/hl_cuda_cudnn.cc
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/hl_cuda_device.cc
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/hl_cudart_wrap.cc
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/hl_dso_loader.cc
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/hl_math.cc
浏览文件 @ e4880016
此差异已折叠。
paddle/cuda/src/hl_time.cc
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/AddtoLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/AgentLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/CRFLayer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/CRFLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/CTCLayer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/CTCLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/CosSimLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/CostLayer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/CostLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/DataLayer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/DataLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/GruCompute.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/Layer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/Layer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/LstmCompute.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/LstmLayer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/LstmLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/MaxLayer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/MaxLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/MixedLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/NCELayer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/NormLayer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/NormLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/Operator.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/Operator.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/PoolLayer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/PoolLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/Projection.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/TensorLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/layers/TransLayer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/tests/TestUtil.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/gserver/tests/TestUtil.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/Allocator.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/BaseMatrix.cu
浏览文件 @ e4880016
此差异已折叠。
paddle/math/BaseMatrix.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/CpuSparseMatrix.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/CpuSparseMatrix.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/ExecViaCpu.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/MathFunctions.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/MathFunctions.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/MathUtils.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/MathUtils.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/Matrix.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/Matrix.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/MatrixBitCode.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/MemoryHandle.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/MemoryHandle.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/PoolAllocator.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/PoolAllocator.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/SIMDFunctions.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/SIMDFunctions.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/SparseMatrix.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/SparseMatrix.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/SparseRowMatrix.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/SparseRowMatrix.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/Storage.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/Vector.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/Vector.h
浏览文件 @ e4880016
此差异已折叠。
paddle/math/tests/test_matrix.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/math/tests/test_matrixUtil.h
浏览文件 @ e4880016
此差异已折叠。
paddle/parameter/Argument.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/parameter/Argument.h
浏览文件 @ e4880016
此差异已折叠。
paddle/parameter/AverageOptimizer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/parameter/Parameter.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/parameter/Parameter.h
浏览文件 @ e4880016
此差异已折叠。
paddle/parameter/Regularizer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/parameter/Regularizer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/parameter/Weight.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/BaseClient.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/BaseClient.h
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/LightNetwork.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/LightNetwork.h
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/ParameterClient2.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/ParameterClient2.h
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/ParameterServer2.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/ParameterServer2.h
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/ProtoServer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/ProtoServer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/RDMANetwork.h
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/SocketChannel.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/SocketChannel.h
浏览文件 @ e4880016
此差异已折叠。
paddle/pserver/test/SocketTest.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/scripts/docker/Dockerfile.m4
浏览文件 @ e4880016
此差异已折叠。
paddle/scripts/submit_local.sh.in
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/ParamUtil.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/ParamUtil.h
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/ParameterUpdater.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/ParameterUpdater.h
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/Tester.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/Tester.h
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/TesterConfig.h
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/Trainer.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/Trainer.h
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/TrainerInternal.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/TrainerInternal.h
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/TrainerMain.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/trainer/tests/picojson.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/BarrierStat.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/BarrierStat.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/ClassRegistrar.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/CommandLineParser.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/CommandLineParser.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/CompilerMacros.h 0 → 100644
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/CustomStackTrace.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/CustomStackTrace.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/DisableCopy.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Excepts.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Flags.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Flags.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/GlobalConstants.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/GlobalConstants.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Locks.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Logging.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Logging.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/PythonUtil.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/PythonUtil.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Queue.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Stat.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/StringUtil.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Thread.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/ThreadLocal.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/ThreadLocal.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/TypeDefs.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Util.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Util.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Version.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/Version.h
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/arch/linux/Locks.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/arch/osx/Locks.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/tests/test_Logging.cpp
浏览文件 @ e4880016
此差异已折叠。
paddle/utils/tests/test_Thread.cpp
浏览文件 @ e4880016
此差异已折叠。
proto/ModelConfig.proto.m4
浏览文件 @ e4880016
此差异已折叠。
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