Merge branch 'develop' of https://github.com/baidu/Paddle into config_parse_bug_fix

avx_docs
dangqingqing 8 years ago
commit 19735cd56e

@ -7,18 +7,14 @@
hooks: hooks:
- id: yapf - id: yapf
- repo: https://github.com/pre-commit/pre-commit-hooks - repo: https://github.com/pre-commit/pre-commit-hooks
sha: 4ef03c4223ad322c7adaa6c6c0efb26b57df3b71 sha: 7539d8bd1a00a3c1bfd34cdb606d3a6372e83469
hooks: hooks:
- id: check-added-large-files - id: check-added-large-files
- id: check-merge-conflict - id: check-merge-conflict
- id: check-symlinks - id: check-symlinks
- id: detect-private-key - id: detect-private-key
- id: end-of-file-fixer - id: end-of-file-fixer
# TODO(yuyang): trailing whitespace has some bugs on markdown - repo: https://github.com/PaddlePaddle/clang-format-pre-commit-hook.git
# files now, please not add it to pre-commit hook now sha: 28c0ea8a67a3e2dbbf4822ef44e85b63a0080a29
# - id: trailing-whitespace hooks:
# - id: clang-formater
# 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

@ -1,10 +1,13 @@
# PaddlePaddle # PaddlePaddle
[![Build Status](https://travis-ci.org/baidu/Paddle.svg?branch=master)](https://travis-ci.org/baidu/Paddle) [![Build Status](https://travis-ci.org/PaddlePaddle/Paddle.svg?branch=develop)](https://travis-ci.org/PaddlePaddle/Paddle)
[![Coverage Status](https://coveralls.io/repos/github/baidu/Paddle/badge.svg?branch=develop)](https://coveralls.io/github/baidu/Paddle?branch=develop) [![Documentation Status](https://img.shields.io/badge/docs-latest-brightgreen.svg?style=flat)](http://www.paddlepaddle.org/)
[![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) [![Documentation Status](https://img.shields.io/badge/中文文档-最新-brightgreen.svg)](http://www.paddlepaddle.org/cn/index.html)
[![License](https://img.shields.io/badge/license-Apache%202.0-green.svg)](LICENSE) [![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. 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. learning to many products at Baidu.
Our vision is to enable deep learning for everyone via PaddlePaddle. 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 ## Features
@ -89,7 +92,7 @@ Both [English Docs](http://paddlepaddle.org/doc/) and [Chinese Docs](http://padd
## Ask Questions ## Ask Questions
You are welcome to submit questions and bug reports as [Github 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 ## Copyright and License
PaddlePaddle is provided under the [Apache-2.0 license](LICENSE). PaddlePaddle is provided under the [Apache-2.0 license](LICENSE).

@ -17,24 +17,15 @@ import os
from optparse import OptionParser from optparse import OptionParser
def extract_dict_features(pair_file, feature_file, src_dict_file, def extract_dict_features(pair_file, feature_file):
tgt_dict_file):
src_dict = set() with open(pair_file) as fin, open(feature_file, 'w') as feature_out:
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:
for line in fin: for line in fin:
sentence, labels = line.strip().split('\t') sentence, predicate, labels = line.strip().split('\t')
sentence_list = sentence.split() sentence_list = sentence.split()
labels_list = labels.split() labels_list = labels.split()
src_dict.update(sentence_list)
tgt_dict.update(labels_list)
verb_index = labels_list.index('B-V') verb_index = labels_list.index('B-V')
verb_feature = sentence_list[verb_index]
mark = [0] * len(labels_list) mark = [0] * len(labels_list)
if verb_index > 0: 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] ctx_n1 = sentence_list[verb_index - 1]
else: else:
ctx_n1 = 'bos' 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 mark[verb_index] = 1
ctx_0_feature = sentence_list[verb_index] ctx_0 = sentence_list[verb_index]
if verb_index < len(labels_list) - 2: if verb_index < len(labels_list) - 2:
mark[verb_index + 1] = 1 mark[verb_index + 1] = 1
ctx_p1 = sentence_list[verb_index + 1] ctx_p1 = sentence_list[verb_index + 1]
else: else:
ctx_p1 = 'eos' 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' \ feature_str = sentence + '\t' \
+ verb_feature + '\t' \ + predicate + '\t' \
+ ctx_n1_feature + '\t' \ + ctx_n2 + '\t' \
+ ctx_0_feature + '\t' \ + ctx_n1 + '\t' \
+ ctx_p1_feature + '\t' \ + ctx_0 + '\t' \
+ ctx_p1 + '\t' \
+ ctx_p2 + '\t' \
+ ' '.join([str(i) for i in mark]) + '\t' \ + ' '.join([str(i) for i in mark]) + '\t' \
+ labels + labels
feature_out.write(feature_str + '\n') 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__': 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 = OptionParser(usage)
parser.add_option('-p', dest='pair_file', help='the pair file') parser.add_option('-p', dest='pair_file', help='the pair file')
parser.add_option( parser.add_option('-f', dest='feature_file', help='the feature file')
'-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')
(options, args) = parser.parse_args() (options, args) = parser.parse_args()
extract_dict_features(options.pair_file, options.feature_file, extract_dict_features(options.pair_file, options.feature_file)
options.src_dict, options.tgt_dict)

@ -51,7 +51,7 @@ def read_sentences(words_file):
for line in fin: for line in fin:
line = line.strip() line = line.strip()
if line == '': if line == '':
sentences.append(s.lower()) sentences.append(s)
s = '' s = ''
else: else:
s += line + ' ' s += line + ' '
@ -64,6 +64,11 @@ def transform_labels(sentences, labels):
if len(labels[i]) == 1: if len(labels[i]) == 1:
continue continue
else: else:
verb_list = []
for x in labels[i][0]:
if x !='-':
verb_list.append(x)
for j in xrange(1, len(labels[i])): for j in xrange(1, len(labels[i])):
label_list = labels[i][j] label_list = labels[i][j]
current_tag = 'O' current_tag = 'O'
@ -88,8 +93,7 @@ def transform_labels(sentences, labels):
is_in_bracket = True is_in_bracket = True
else: else:
print 'error:', ll print 'error:', ll
sen_lab_pair.append((sentences[i], verb_list[j-1], label_seq))
sen_lab_pair.append((sentences[i], label_seq))
return sen_lab_pair return sen_lab_pair
@ -97,9 +101,9 @@ def write_file(sen_lab_pair, output_file):
with open(output_file, 'w') as fout: with open(output_file, 'w') as fout:
for x in sen_lab_pair: for x in sen_lab_pair:
sentence = x[0] sentence = x[0]
label_seq = ' '.join(x[1]) label_seq = ' '.join(x[2])
assert len(sentence.split()) == len(x[1]) assert len(sentence.split()) == len(x[2])
fout.write(sentence + '\t' + label_seq + '\n') fout.write(sentence + '\t' + x[1]+'\t' +label_seq + '\n')
if __name__ == '__main__': if __name__ == '__main__':

@ -14,6 +14,10 @@
# limitations under the License. # limitations under the License.
set -e set -e
wget http://www.cs.upc.edu/~srlconll/conll05st-tests.tar.gz 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 tar -xzvf conll05st-tests.tar.gz
rm conll05st-tests.tar.gz rm conll05st-tests.tar.gz
cp ./conll05st-release/test.wsj/words/test.wsj.words.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 gunzip test.wsj.props.gz
python extract_pairs.py -w test.wsj.words -p test.wsj.props -o test.wsj.seq_pair 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

@ -17,11 +17,15 @@ from paddle.trainer.PyDataProvider2 import *
UNK_IDX = 0 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.word_dict = word_dict
settings.label_dict = label_dict settings.label_dict = label_dict
settings.predicate_dict = predicate_dict
#all inputs are integral and sequential type #all inputs are integral and sequential type
settings.slots = [ 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(len(word_dict)),
@ -31,27 +35,33 @@ def hook(settings, word_dict, label_dict, **kwargs):
] ]
@provider(init_hook=hook) def get_batch_size(yeild_data):
def process(obj, file_name): 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: with open(file_name, 'r') as fdata:
for line in 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') line.strip().split('\t')
words = sentence.split() words = sentence.split()
sen_len = len(words) 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 predicate_slot = [settings.predicate_dict.get(predicate)] * sen_len
ctx_n1_slot = [obj.word_dict.get(ctx_n1, UNK_IDX)] * sen_len ctx_n2_slot = [settings.word_dict.get(ctx_n2, UNK_IDX)] * sen_len
ctx_0_slot = [obj.word_dict.get(ctx_0, UNK_IDX)] * sen_len ctx_n1_slot = [settings.word_dict.get(ctx_n1, UNK_IDX)] * sen_len
ctx_p1_slot = [obj.word_dict.get(ctx_p1, 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() marks = mark.split()
mark_slot = [int(w) for w in marks] mark_slot = [int(w) for w in marks]
label_list = label.split() label_list = label.split()
label_slot = [obj.label_dict.get(w) for w in label_list] label_slot = [settings.label_dict.get(w) for w in label_list]
yield word_slot, predicate_slot, ctx_n2_slot, ctx_n1_slot, \
yield word_slot, predicate_slot, ctx_n1_slot, \ ctx_0_slot, ctx_p1_slot, ctx_p2_slot, mark_slot, label_slot
ctx_0_slot, ctx_p1_slot, mark_slot, label_slot

@ -18,8 +18,9 @@ import sys
from paddle.trainer_config_helpers import * from paddle.trainer_config_helpers import *
#file paths #file paths
word_dict_file = './data/src.dict' word_dict_file = './data/wordDict.txt'
label_dict_file = './data/tgt.dict' label_dict_file = './data/targetDict.txt'
predicate_file= './data/verbDict.txt'
train_list_file = './data/train.list' train_list_file = './data/train.list'
test_list_file = './data/test.list' test_list_file = './data/test.list'
@ -30,8 +31,10 @@ if not is_predict:
#load dictionaries #load dictionaries
word_dict = dict() word_dict = dict()
label_dict = dict() label_dict = dict()
predicate_dict = dict()
with open(word_dict_file, 'r') as f_word, \ 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): for i, line in enumerate(f_word):
w = line.strip() w = line.strip()
word_dict[w] = i word_dict[w] = i
@ -40,6 +43,11 @@ if not is_predict:
w = line.strip() w = line.strip()
label_dict[w] = i label_dict[w] = i
for i, line in enumerate(f_pre):
w = line.strip()
predicate_dict[w] = i
if is_test: if is_test:
train_list_file = None train_list_file = None
@ -50,91 +58,157 @@ if not is_predict:
module='dataprovider', module='dataprovider',
obj='process', obj='process',
args={'word_dict': word_dict, args={'word_dict': word_dict,
'label_dict': label_dict}) 'label_dict': label_dict,
'predicate_dict': predicate_dict })
word_dict_len = len(word_dict) word_dict_len = len(word_dict)
label_dict_len = len(label_dict) label_dict_len = len(label_dict)
pred_len = len(predicate_dict)
else: else:
word_dict_len = get_config_arg('dict_len', int) word_dict_len = get_config_arg('dict_len', int)
label_dict_len = get_config_arg('label_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 mark_dict_len = 2
word_dim = 32 word_dim = 32
mark_dim = 5 mark_dim = 5
hidden_dim = 128 hidden_dim = 512
depth = 8 depth = 8
emb_lr = 1e-2
fc_lr = 1e-2
lstm_lr = 2e-2
########################### Optimizer #######################################
settings( settings(
batch_size=150, batch_size=150,
learning_method=AdamOptimizer(), learning_method=MomentumOptimizer(momentum=0),
learning_rate=1e-3, learning_rate=2e-2,
regularization=L2Regularization(8e-4), 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) 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_n1 = data_layer(name='ctx_n1_data', size=word_dict_len)
ctx_0 = data_layer(name='ctx_0_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_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) mark = data_layer(name='mark_data', size=mark_dict_len)
if not is_predict: if not is_predict:
target = data_layer(name='target', size=label_dict_len) 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) default_std=1/math.sqrt(hidden_dim)/3.0
predicate_embedding = embedding_layer(
size=word_dim, input=predicate, param_attr=ptt) emb_para = ParameterAttribute(name='emb', initial_std=0., learning_rate=0.)
ctx_n1_embedding = embedding_layer(size=word_dim, input=ctx_n1, param_attr=ptt) std_0 = ParameterAttribute(initial_std=0.)
ctx_0_embedding = embedding_layer(size=word_dim, input=ctx_0, param_attr=ptt) std_default = ParameterAttribute(initial_std=default_std)
ctx_p1_embedding = embedding_layer(size=word_dim, input=ctx_p1, param_attr=ptt)
mark_embedding = embedding_layer(size=mark_dim, input=mark) 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( hidden_0 = mixed_layer(
name='hidden0',
size=hidden_dim, size=hidden_dim,
input=[ bias_attr=std_default,
full_matrix_projection(input=word_embedding), input=[ full_matrix_projection(input=emb, param_attr=std_default ) for emb in emb_layers ])
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),
])
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 #stack L-LSTM and R-LSTM with direct edges
input_tmp = [hidden_0, lstm_0] input_tmp = [hidden_0, lstm_0]
for i in range(1, depth): 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: if not is_predict:
cls = classification_cost(input=prob, label=target) crf_l = crf_layer( name = 'crf',
outputs(cls) 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: 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)

@ -26,7 +26,7 @@ UNK_IDX = 0
class Prediction(): 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. train_conf: trainer configure.
dict_file: word dictionary file name. dict_file: word dictionary file name.
@ -35,26 +35,41 @@ class Prediction():
self.dict = {} self.dict = {}
self.labels = {} self.labels = {}
self.predicate_dict={}
self.labels_reverse = {} 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_dict = len(self.dict)
len_label = len(self.labels) len_label = len(self.labels)
len_pred = len(self.predicate_dict)
conf = parse_config(train_conf, 'dict_len=' + str(len_dict) +
',label_len=' + str(len_label) + ',is_predict=True') 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( self.network = swig_paddle.GradientMachine.createFromConfigProto(
conf.model_config) conf.model_config)
self.network.loadParameters(model_dir) self.network.loadParameters(model_dir)
slots = [ 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(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(2)
] ]
self.converter = DataProviderConverter(slots) 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. Load dictionary from self.dict_file.
""" """
@ -65,39 +80,42 @@ class Prediction():
self.labels[line.strip()] = line_count self.labels[line.strip()] = line_count
self.labels_reverse[line_count] = line.strip() 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): def get_data(self, data_file):
""" """
Get input data of paddle format. Get input data of paddle format.
""" """
with open(data_file, 'r') as fdata: with open(data_file, 'r') as fdata:
for line in 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') ).split('\t')
words = sentence.split() words = sentence.split()
sen_len = len(words) sen_len = len(words)
word_slot = [self.dict.get(w, UNK_IDX) for w in 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_n1_slot = [self.dict.get(ctx_n1, UNK_IDX)] * sen_len
ctx_0_slot = [self.dict.get(ctx_0, 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_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() marks = mark.split()
mark_slot = [int(w) for w in marks] 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, \ def predict(self, data_file, output_file):
ctx_0_slot, ctx_p1_slot, mark_slot
def predict(self, data_file):
""" """
data_file: file name of input data. data_file: file name of input data.
""" """
input = self.converter(self.get_data(data_file)) input = self.converter(self.get_data(data_file))
output = self.network.forwardTest(input) output = self.network.forwardTest(input)
prob = output[0]["value"] lab = output[0]["id"].tolist()
lab = list(np.argsort(-prob)[:, 0])
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 index = 0
for line in fin: for line in fin:
sen = line.split('\t')[0] sen = line.split('\t')[0]
@ -109,8 +127,8 @@ class Prediction():
def option_parser(): def option_parser():
usage = ("python predict.py -c config -w model_dir " usage = ("python predict.py -c config -w model_dir "
"-d word dictionary -l label_file -i input_file") "-d word dictionary -l label_file -i input_file -p pred_dict_file")
parser = OptionParser(usage="usage: %s [options]" % usage) parser = OptionParser(usage="usage: %s [options]" % usage)
parser.add_option( parser.add_option(
"-c", "-c",
@ -131,6 +149,13 @@ def option_parser():
dest="label_file", dest="label_file",
default=None, default=None,
help="label file") 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( parser.add_option(
"-i", "-i",
"--data", "--data",
@ -144,6 +169,14 @@ def option_parser():
dest="model_path", dest="model_path",
default=None, default=None,
help="model path") help="model path")
parser.add_option(
"-o",
"--output_file",
action="store",
dest="output_file",
default=None,
help="output file")
return parser.parse_args() return parser.parse_args()
@ -154,10 +187,12 @@ def main():
dict_file = options.dict_file dict_file = options.dict_file
model_path = options.model_path model_path = options.model_path
label_file = options.label_file label_file = options.label_file
predict_dict_file = options.predict_dict_file
output_file = options.output_file
swig_paddle.initPaddle("--use_gpu=0") swig_paddle.initPaddle("--use_gpu=0")
predict = Prediction(train_conf, dict_file, model_path, label_file) predict = Prediction(train_conf, dict_file, model_path, label_file, predict_dict_file)
predict.predict(data_file) predict.predict(data_file,output_file)
if __name__ == '__main__': if __name__ == '__main__':

@ -26,15 +26,18 @@ LOG=`get_best_pass $log`
LOG=(${LOG}) LOG=(${LOG})
best_model_path="output/pass-${LOG[1]}" best_model_path="output/pass-${LOG[1]}"
config_file=db_lstm.py config_file=db_lstm.py
dict_file=./data/src.dict dict_file=./data/wordDict.txt
label_file=./data/tgt.dict label_file=./data/targetDict.txt
predicate_dict_file=./data/verbDict.txt
input_file=./data/feature input_file=./data/feature
output_file=predict.res
python predict.py \ python predict.py \
-c $config_file \ -c $config_file \
-w $best_model_path \ -w $best_model_path \
-l $label_file \ -l $label_file \
-p $predicate_dict_file \
-d $dict_file \ -d $dict_file \
-i $input_file -i $input_file \
-o $output_file

@ -36,4 +36,5 @@ paddle train \
--job=test \ --job=test \
--use_gpu=false \ --use_gpu=false \
--config_args=is_test=1 \ --config_args=is_test=1 \
--test_all_data_in_one_period=1 \
2>&1 | tee 'test.log' 2>&1 | tee 'test.log'

@ -16,11 +16,14 @@
set -e set -e
paddle train \ paddle train \
--config=./db_lstm.py \ --config=./db_lstm.py \
--use_gpu=0 \
--log_period=5000 \
--trainer_count=1 \
--show_parameter_stats_period=5000 \
--save_dir=./output \ --save_dir=./output \
--trainer_count=4 \ --num_passes=10000 \
--log_period=10 \ --average_test_period=10000000 \
--num_passes=500 \ --init_model_path=./data \
--use_gpu=false \ --load_missing_parameter_strategy=rand \
--show_parameter_stats_period=10 \
--test_all_data_in_one_period=1 \ --test_all_data_in_one_period=1 \
2>&1 | tee 'train.log' 2>&1 | tee 'train.log'

Binary file not shown.

After

Width:  |  Height:  |  Size: 52 KiB

File diff suppressed because it is too large Load Diff

@ -81,5 +81,8 @@ else()
add_library(paddle_cuda ${CUDA_SOURCES}) add_library(paddle_cuda ${CUDA_SOURCES})
endif() endif()
add_style_check_target(paddle_cuda ${CUDA_SOURCES}) add_style_check_target(paddle_cuda
add_style_check_target(paddle_cuda ${CUDA_HEADERS}) ${CUDA_SOURCES}
${CUDA_HEADERS}
${CUDA_DSO_SOURCES}
${CUDA_CXX_WITH_GPU_SOURCES})

@ -104,7 +104,7 @@ CUBLAS_BLAS_ROUTINE_EACH(DYNAMIC_LOAD_CUBLAS_V2_WRAP)
#endif #endif
const char* hl_cublas_get_error_string(cublasStatus_t status) { const char* hl_cublas_get_error_string(cublasStatus_t status) {
switch(status) { switch (status) {
case CUBLAS_STATUS_NOT_INITIALIZED: case CUBLAS_STATUS_NOT_INITIALIZED:
return "[cublas status]: not initialized"; return "[cublas status]: not initialized";
case CUBLAS_STATUS_ALLOC_FAILED: case CUBLAS_STATUS_ALLOC_FAILED:
@ -181,7 +181,7 @@ void hl_matrix_inverse(real *A_d, real *C_d, int dimN, int lda, int ldc) {
real **inout_d = (real **)hl_malloc_device(sizeof(real *)); real **inout_d = (real **)hl_malloc_device(sizeof(real *));
hl_memcpy(inout_d, inout_h, sizeof(real *)); hl_memcpy(inout_d, inout_h, sizeof(real *));
int *pivot_d = (int *)hl_malloc_device(dimN*sizeof(int)); int *pivot_d = (int *)hl_malloc_device(dimN * sizeof(int));
int *info_d = (int *)t_resource.gpu_mem; int *info_d = (int *)t_resource.gpu_mem;
/* Note: cublasSgetrfBatched is used to calculate a number of /* Note: cublasSgetrfBatched is used to calculate a number of
@ -189,10 +189,9 @@ void hl_matrix_inverse(real *A_d, real *C_d, int dimN, int lda, int ldc) {
the API for better performance. the API for better performance.
*/ */
CHECK_CUBLAS(CUBLAS_GETRF(t_resource.handle, CHECK_CUBLAS(CUBLAS_GETRF(t_resource.handle,
dimN, inout_d, lda, pivot_d, dimN, inout_d, lda, pivot_d, info_d, 1));
info_d, 1));
int info_h; int info_h;
hl_memcpy(&info_h, info_d, sizeof(int)); hl_memcpy(&info_h, info_d, sizeof(int));
if (info_h != 0) { if (info_h != 0) {
LOG(FATAL) << "Factorization of matrix failed: matrix may be singular.\n"; LOG(FATAL) << "Factorization of matrix failed: matrix may be singular.\n";
@ -204,8 +203,8 @@ void hl_matrix_inverse(real *A_d, real *C_d, int dimN, int lda, int ldc) {
hl_memcpy(out_d, out_h, sizeof(real *)); hl_memcpy(out_d, out_h, sizeof(real *));
CHECK_CUBLAS(CUBLAS_GETRI(t_resource.handle, CHECK_CUBLAS(CUBLAS_GETRI(t_resource.handle,
dimN, (const real **)inout_d, lda, pivot_d, dimN, (const real **)inout_d, lda, pivot_d,
out_d, ldc, info_d, 1)); out_d, ldc, info_d, 1));
hl_memcpy(&info_h, info_d, sizeof(int)); hl_memcpy(&info_h, info_d, sizeof(int));
if (info_h != 0) { if (info_h != 0) {
@ -215,7 +214,7 @@ void hl_matrix_inverse(real *A_d, real *C_d, int dimN, int lda, int ldc) {
hl_free_mem_device(inout_d); hl_free_mem_device(inout_d);
hl_free_mem_device(pivot_d); hl_free_mem_device(pivot_d);
hl_free_mem_device(out_d); hl_free_mem_device(out_d);
CHECK_SYNC("hl_matrix_inverse failed"); CHECK_SYNC("hl_matrix_inverse failed");
} }

File diff suppressed because it is too large Load Diff

@ -203,8 +203,8 @@ inline pid_t gettid() {
#endif #endif
pid_t tid = syscall(__NR_gettid); pid_t tid = syscall(__NR_gettid);
#endif #endif
CHECK_NE(tid, -1); CHECK_NE((int)tid, -1);
return tid; return tid;
} }
void hl_init(int device) { void hl_init(int device) {
@ -355,7 +355,8 @@ void* hl_malloc_host(size_t size) {
void *dest_h; void *dest_h;
CHECK(size) << __func__ << ": the size for device memory is 0, please check."; CHECK(size) << __func__ << ": the size for device memory is 0, please check.";
CHECK_CUDA(dynload::cudaHostAlloc((void**)&dest_h, size, cudaHostAllocDefault)); CHECK_CUDA(dynload::cudaHostAlloc(
(void**)&dest_h, size, cudaHostAllocDefault));
return dest_h; return dest_h;
} }
@ -364,7 +365,7 @@ void hl_free_mem_host(void *dest_h) {
CHECK_NOTNULL(dest_h); CHECK_NOTNULL(dest_h);
cudaError_t err = dynload::cudaFreeHost(dest_h); cudaError_t err = dynload::cudaFreeHost(dest_h);
CHECK (cudaSuccess == err || cudaErrorCudartUnloading == err) CHECK(cudaSuccess == err || cudaErrorCudartUnloading == err)
<< hl_get_device_error_string(); << hl_get_device_error_string();
} }
@ -502,7 +503,8 @@ int hl_get_cuda_version() {
return g_cuda_lib_version; return g_cuda_lib_version;
} }
void hl_create_thread_resources(int device, thread_device_resources device_res) { void hl_create_thread_resources(int device,
thread_device_resources device_res) {
CHECK_CUDA(dynload::cudaSetDevice(device)); CHECK_CUDA(dynload::cudaSetDevice(device));
/* create thread stream */ /* create thread stream */

@ -78,48 +78,38 @@ __host__ cudaError_t CUDARTAPI cudaLaunchKernel(const void *func,
dim3 blockDim, dim3 blockDim,
void **args, void **args,
size_t sharedMem, size_t sharedMem,
cudaStream_t stream) cudaStream_t stream) {
{ return dynload::cudaLaunchKernel(func, gridDim, blockDim,
return dynload::cudaLaunchKernel(func, gridDim, blockDim, args, sharedMem, stream); args, sharedMem, stream);
} }
#endif /* CUDART_VERSION >= 7000 */ #endif /* CUDART_VERSION >= 7000 */
__host__ cudaError_t CUDARTAPI cudaLaunch(const void *func) __host__ cudaError_t CUDARTAPI cudaLaunch(const void *func) {
{
return dynload::cudaLaunch(func); return dynload::cudaLaunch(func);
} }
__host__ cudaError_t CUDARTAPI cudaSetupArgument(const void *arg, __host__ cudaError_t CUDARTAPI cudaSetupArgument(const void *arg,
size_t size, size_t size,
size_t offset) size_t offset) {
{
return dynload::cudaSetupArgument(arg, size, offset); return dynload::cudaSetupArgument(arg, size, offset);
} }
__host__ cudaError_t CUDARTAPI cudaConfigureCall(dim3 gridDim, __host__ cudaError_t CUDARTAPI cudaConfigureCall(dim3 gridDim,
dim3 blockDim, dim3 blockDim,
size_t sharedMem, size_t sharedMem,
cudaStream_t stream) cudaStream_t stream) {
{
return dynload::cudaConfigureCall(gridDim, blockDim, return dynload::cudaConfigureCall(gridDim, blockDim,
sharedMem, stream); sharedMem, stream);
} }
extern "C" { extern "C" {
void** CUDARTAPI __cudaRegisterFatBinary( void** CUDARTAPI __cudaRegisterFatBinary(void *fatCubin) {
void *fatCubin
)
{
return dynload::__cudaRegisterFatBinary(fatCubin); return dynload::__cudaRegisterFatBinary(fatCubin);
} }
void CUDARTAPI __cudaUnregisterFatBinary( void CUDARTAPI __cudaUnregisterFatBinary(void **fatCubinHandle) {
void **fatCubinHandle
)
{
return dynload::__cudaUnregisterFatBinary(fatCubinHandle); return dynload::__cudaUnregisterFatBinary(fatCubinHandle);
} }

@ -12,27 +12,28 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#include "hl_dso_loader.h" #include "hl_dso_loader.h"
#include "paddle/utils/Logging.h"
#include "paddle/utils/CommandLineParser.h" #include "paddle/utils/CommandLineParser.h"
#include "paddle/utils/Logging.h"
P_DEFINE_string(cudnn_dir, "", P_DEFINE_string(cudnn_dir,
"",
"Specify path for loading libcudnn.so. For instance, " "Specify path for loading libcudnn.so. For instance, "
"/usr/local/cudnn/lib64. If empty [default], dlopen will search " "/usr/local/cudnn/lib. If empty [default], dlopen "
"cudnn from LD_LIBRARY_PATH"); "will search cudnn from LD_LIBRARY_PATH");
P_DEFINE_string(cuda_dir, "", P_DEFINE_string(cuda_dir,
"",
"Specify path for loading cuda library, such as libcublas, " "Specify path for loading cuda library, such as libcublas, "
"libcurand. For instance, /usr/local/cuda/lib64. " "libcurand. For instance, /usr/local/cuda/lib64. (Note: "
"(Note: libcudart can not be specified by cuda_dir, since some " "libcudart can not be specified by cuda_dir, since some "
"build-in function in cudart already ran before main entry). " "build-in function in cudart already ran before main entry). "
"If empty [default], dlopen will search cuda from LD_LIBRARY_PATH"); "If default, dlopen will search cuda from LD_LIBRARY_PATH");
static inline std::string join(const std::string& part1, const std::string& part2) { static inline std::string join(const std::string& part1,
const std::string& part2) {
// directory separator // directory separator
const char sep = '/'; const char sep = '/';
if (!part2.empty() && part2.front() == sep) { if (!part2.empty() && part2.front() == sep) {
return part2; return part2;
} }
@ -46,100 +47,115 @@ static inline std::string join(const std::string& part1, const std::string& part
return ret; return ret;
} }
static inline void GetDsoHandleFromDefaultPath( static inline void GetDsoHandleFromDefaultPath(std::string& dso_path,
std::string& dso_path, void** dso_handle, int dynload_flags) { void** dso_handle,
VLOG(3) << "Try to find cuda library: " << dso_path int dynload_flags) {
<< " from default system path."; VLOG(3) << "Try to find cuda library: " << dso_path
// default search from LD_LIBRARY_PATH/DYLD_LIBRARY_PATH << " from default system path.";
// default search from LD_LIBRARY_PATH/DYLD_LIBRARY_PATH
*dso_handle = dlopen(dso_path.c_str(), dynload_flags);
// DYLD_LIBRARY_PATH is disabled after Mac OS 10.11 to
// bring System Integrity Projection (SIP), if dso_handle
// is null, search from default package path in Mac OS.
#if defined(__APPLE__) || defined(__OSX__)
if (nullptr == *dso_handle) {
dso_path = join("/usr/local/cuda/lib/", dso_path);
*dso_handle = dlopen(dso_path.c_str(), dynload_flags); *dso_handle = dlopen(dso_path.c_str(), dynload_flags);
// DYLD_LIBRARY_PATH is disabled after Mac OS 10.11 to
// bring System Integrity Projection (SIP), if dso_handle
// is null, search from default package path in Mac OS.
#if defined(__APPLE__) || defined(__OSX__)
if (nullptr == *dso_handle) { if (nullptr == *dso_handle) {
dso_path = join("/usr/local/cuda/lib/", dso_path); if (dso_path == "libcudnn.dylib") {
*dso_handle = dlopen(dso_path.c_str(), dynload_flags); LOG(FATAL)
if (nullptr == *dso_handle) { << "Note: [Recommend] copy cudnn into /usr/local/cuda/ \n" // NOLINT
if (dso_path == "libcudnn.dylib") { << "For instance, sudo tar -xzf "
LOG(FATAL) << "Note: [Recommend] copy cudnn into /usr/local/cuda/ \n" "cudnn-7.5-osx-x64-v5.0-ga.tgz -C " // NOLINT
<< "For instance, sudo tar -xzf cudnn-7.5-osx-x64-v5.0-ga.tgz -C " << "/usr/local \n sudo chmod a+r "
<< "/usr/local \n sudo chmod a+r /usr/local/cuda/include/cudnn.h " "/usr/local/cuda/include/cudnn.h " // NOLINT
<< "/usr/local/cuda/lib/libcudnn*"; << "/usr/local/cuda/lib/libcudnn*";
} }
} }
} }
#endif #endif
} }
static inline void GetDsoHandleFromSearchPath( static inline void GetDsoHandleFromSearchPath(const std::string& search_root,
const std::string& search_root, const std::string& dso_name,
const std::string& dso_name, void** dso_handle) {
void** dso_handle) { int dynload_flags = RTLD_LAZY | RTLD_LOCAL;
int dynload_flags = RTLD_LAZY | RTLD_LOCAL; *dso_handle = nullptr;
*dso_handle = nullptr;
std::string dlPath = dso_name;
std::string dlPath = dso_name; if (search_root.empty()) {
if (search_root.empty()) { GetDsoHandleFromDefaultPath(dlPath, dso_handle, dynload_flags);
GetDsoHandleFromDefaultPath(dlPath, dso_handle, dynload_flags); } else {
} else { // search xxx.so from custom path
// search xxx.so from custom path dlPath = join(search_root, dso_name);
dlPath = join(search_root, dso_name); *dso_handle = dlopen(dlPath.c_str(), dynload_flags);
*dso_handle = dlopen(dlPath.c_str(), dynload_flags); // if not found, search from default path
// if not found, search from default path if (nullptr == *dso_handle) {
if (nullptr == dso_handle) { LOG(WARNING) << "Failed to find cuda library: " << dlPath;
LOG(WARNING) << "Failed to find cuda library: " << dlPath; dlPath = dso_name;
dlPath = dso_name; GetDsoHandleFromDefaultPath(dlPath, dso_handle, dynload_flags);
GetDsoHandleFromDefaultPath(dlPath, dso_handle, dynload_flags);
}
} }
}
CHECK(nullptr != *dso_handle) CHECK(nullptr != *dso_handle) << "Failed to find cuda library: " << dlPath
<< "Failed to find cuda library: " << dlPath << std::endl << std::endl
<< "Please specify its path correctly using one of the following ideas: \n" << "Please specify its path correctly using "
"one of the following ways: \n" // NOLINT
<< "Idea 1. set cuda and cudnn lib path at runtime. "
<< "http://www.paddlepaddle.org/doc/ui/cmd_argument/argument_outline.html \n" << "Method 1. set cuda and cudnn lib path at "
<< "For instance, issue command: paddle train --use_gpu=1 " "runtime. "
<< "--cuda_dir=/usr/local/cudnn/lib --cudnn_dir=/usr/local/cudnn/lib ...\n" << "http://www.paddlepaddle.org/doc/ui/"
"cmd_argument/"
<< "Idea 2. set environment variable LD_LIBRARY_PATH on Linux or " "argument_outline.html \n" // NOLINT
<< "DYLD_LIBRARY_PATH on Mac OS. \n" << "For instance, issue command: paddle train "
<< "For instance, issue command: export LD_LIBRARY_PATH=... \n" "--use_gpu=1 "
<< "--cuda_dir=/usr/local/cuda/lib64 "
<< "Note: After Mac OS 10.11, using the DYLD_LIBRARY_PATH is impossible " "--cudnn_dir=/usr/local/cudnn/lib "
<< "unless System Integrity Protection (SIP) is disabled. However, @Idea 1" "...\n" // NOLINT
<< "always work well.";
<< "Method 2. set environment variable "
"LD_LIBRARY_PATH on Linux or "
<< "DYLD_LIBRARY_PATH on Mac OS. \n"
<< "For instance, issue command: export "
"LD_LIBRARY_PATH=... \n"
<< "Note: After Mac OS 10.11, using the "
"DYLD_LIBRARY_PATH is impossible "
<< "unless System Integrity Protection (SIP) "
"is disabled. However, "
"method 1 " // NOLINT
<< "always work well.";
} }
void GetCublasDsoHandle(void** dso_handle) { void GetCublasDsoHandle(void** dso_handle) {
#if defined(__APPLE__) || defined(__OSX__) #if defined(__APPLE__) || defined(__OSX__)
GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcublas.dylib", dso_handle); GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcublas.dylib", dso_handle);
#else #else
GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcublas.so", dso_handle); GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcublas.so", dso_handle);
#endif #endif
} }
void GetCudnnDsoHandle(void** dso_handle) { void GetCudnnDsoHandle(void** dso_handle) {
#if defined(__APPLE__) || defined(__OSX__) #if defined(__APPLE__) || defined(__OSX__)
GetDsoHandleFromSearchPath(FLAGS_cudnn_dir, "libcudnn.dylib", dso_handle); GetDsoHandleFromSearchPath(FLAGS_cudnn_dir, "libcudnn.dylib", dso_handle);
#else #else
GetDsoHandleFromSearchPath(FLAGS_cudnn_dir, "libcudnn.so", dso_handle); GetDsoHandleFromSearchPath(FLAGS_cudnn_dir, "libcudnn.so", dso_handle);
#endif #endif
} }
void GetCudartDsoHandle(void** dso_handle) { void GetCudartDsoHandle(void** dso_handle) {
#if defined(__APPLE__) || defined(__OSX__) #if defined(__APPLE__) || defined(__OSX__)
GetDsoHandleFromSearchPath("", "libcudart.dylib", dso_handle); GetDsoHandleFromSearchPath("", "libcudart.dylib", dso_handle);
#else #else
GetDsoHandleFromSearchPath("", "libcudart.so", dso_handle); GetDsoHandleFromSearchPath("", "libcudart.so", dso_handle);
#endif #endif
} }
void GetCurandDsoHandle(void** dso_handle) { void GetCurandDsoHandle(void** dso_handle) {
#if defined(__APPLE__) || defined(__OSX__) #if defined(__APPLE__) || defined(__OSX__)
GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcurand.dylib", dso_handle); GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcurand.dylib", dso_handle);
#else #else
GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcurand.so", dso_handle); GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcurand.so", dso_handle);
#endif #endif
} }

@ -240,7 +240,7 @@ public:
seqClassficationError_ = 0; seqClassficationError_ = 0;
} }
virtual void printStats(std::ostream& os) { virtual void printStats(std::ostream& os) const {
os << config_.name() << "=" os << config_.name() << "="
<< (numSequences_ ? totalScore_ / numSequences_ : 0); << (numSequences_ ? totalScore_ / numSequences_ : 0);
os << " deletions error" os << " deletions error"

@ -114,7 +114,7 @@ public:
numCorrect_ = 0; numCorrect_ = 0;
} }
virtual void printStats(std::ostream& os) { virtual void printStats(std::ostream& os) const {
double precision = (double)numCorrect_ / numOutputSegments_; double precision = (double)numCorrect_ / numOutputSegments_;
double recall = (double)numCorrect_ / numLabelSegments_; double recall = (double)numCorrect_ / numLabelSegments_;
double f1 = double f1 =

@ -315,7 +315,7 @@ public:
return 0; return 0;
} }
virtual void printStats(std::ostream& os) { virtual void printStats(std::ostream& os) const {
CHECK(colIdx_ + (int32_t)colNum_ >= 0 && colIdx_ - (int32_t)colNum_ < 0) CHECK(colIdx_ + (int32_t)colNum_ >= 0 && colIdx_ - (int32_t)colNum_ < 0)
<< "column index [" << colIdx_ << "] out of range [-" << colNum_ << ", " << "column index [" << colIdx_ << "] out of range [-" << colNum_ << ", "
<< colNum_ << ")"; << colNum_ << ")";
@ -421,7 +421,7 @@ void AucEvaluator::distributeEval(ParameterClient2* client) {
client->reduce(statNeg_, statNeg_, kBinNum_ + 1, FLAGS_trainer_id, 0); client->reduce(statNeg_, statNeg_, kBinNum_ + 1, FLAGS_trainer_id, 0);
} }
double AucEvaluator::calcAuc() { double AucEvaluator::calcAuc() const {
double totPos = 0.0; double totPos = 0.0;
double totNeg = 0.0; double totNeg = 0.0;
double totPosPrev = 0.0; double totPosPrev = 0.0;
@ -584,7 +584,7 @@ real PrecisionRecallEvaluator::evalImp(std::vector<Argument>& arguments) {
return 0; return 0;
} }
void PrecisionRecallEvaluator::printStats(std::ostream& os) { void PrecisionRecallEvaluator::printStats(std::ostream& os) const {
int label = config_.positive_label(); int label = config_.positive_label();
if (label != -1) { if (label != -1) {
CHECK(label >= 0 && label < (int)statsInfo_.size()) CHECK(label >= 0 && label < (int)statsInfo_.size())

@ -99,19 +99,19 @@ public:
* @brief print the statistics of evaluate result * @brief print the statistics of evaluate result
* @note finish() should be called before printStats * @note finish() should be called before printStats
*/ */
virtual void printStats(std::ostream& os) { virtual void printStats(std::ostream& os) const {
os << config_.name() << "=" os << config_.name() << "="
<< (numSamples_ ? totalScore_ / numSamples_ : 0); << (numSamples_ ? totalScore_ / numSamples_ : 0);
} }
friend std::ostream& operator<<(std::ostream& os, friend std::ostream& operator<<(std::ostream& os,
Evaluator& evaluator) { const Evaluator& evaluator) {
evaluator.printStats(os); evaluator.printStats(os);
return os; return os;
} }
friend std::ostream&& operator<<(std::ostream&& os, // NOLINT friend std::ostream&& operator<<(std::ostream&& os, // NOLINT
Evaluator& evaluator) { const Evaluator& evaluator) {
evaluator.printStats(os); evaluator.printStats(os);
return std::move(os); return std::move(os);
} }
@ -135,7 +135,7 @@ public:
return -1; return -1;
} }
virtual void finish() {} virtual void finish() {}
virtual void printStats(std::ostream&) {} virtual void printStats(std::ostream&) const {}
}; };
/** /**
* @brief evaluate AUC using colIdx-th column as prediction. * @brief evaluate AUC using colIdx-th column as prediction.
@ -165,7 +165,7 @@ public:
virtual real evalImp(std::vector<Argument>& arguments); virtual real evalImp(std::vector<Argument>& arguments);
virtual void printStats(std::ostream& os) { virtual void printStats(std::ostream& os) const {
os << config_.name() << "=" << calcAuc(); os << config_.name() << "=" << calcAuc();
} }
@ -189,7 +189,7 @@ private:
return (X1 > X2 ? (X1 - X2) : (X2 - X1)) * (Y1 + Y2) / 2.0; return (X1 > X2 ? (X1 - X2) : (X2 - X1)) * (Y1 + Y2) / 2.0;
} }
double calcAuc(); double calcAuc() const;
}; };
/** /**
@ -244,7 +244,7 @@ public:
virtual real evalImp(std::vector<Argument>& arguments); virtual real evalImp(std::vector<Argument>& arguments);
virtual void printStats(std::ostream& os); virtual void printStats(std::ostream& os) const;
virtual void distributeEval(ParameterClient2* client); virtual void distributeEval(ParameterClient2* client);
@ -339,7 +339,7 @@ public:
virtual void finish() { calc(predictArray_); } virtual void finish() { calc(predictArray_); }
virtual void printStats(std::ostream& os) { virtual void printStats(std::ostream& os) const {
os << " pos/neg" os << " pos/neg"
<< "=" << pairArray_[0] / ((pairArray_[1] <= 0) ? 1.0 : pairArray_[1]); << "=" << pairArray_[0] / ((pairArray_[1] <= 0) ? 1.0 : pairArray_[1]);
} }

@ -154,7 +154,7 @@ public:
return -1; return -1;
} }
virtual void printStats(std::ostream& os) { virtual void printStats(std::ostream& os) const {
for (auto& evaluator : evaluators_) { for (auto& evaluator : evaluators_) {
evaluator->printStats(os); evaluator->printStats(os);
os << ' '; os << ' ';

@ -325,7 +325,7 @@ public:
(void)arguments; (void)arguments;
return -1; return -1;
} }
virtual void printStats(std::ostream& os) { virtual void printStats(std::ostream& os) const {
for (auto& evaluator : evaluators_) { for (auto& evaluator : evaluators_) {
evaluator->printStats(os); evaluator->printStats(os);
os << ' '; os << ' ';

Some files were not shown because too many files have changed in this diff Show More

Loading…
Cancel
Save