Skip to content

M
models

PaddlePaddle / models
大约 2 年 前同步成功

通知 232
Star 6828
Fork 2962
M
models
未验证 提交 7d338bb0 编写于 作者: Y Yibing Liu 提交者： GitHub
浏览文件
操作

Remove dirs unarchived & legacy in release 1.6 (#3666)

上级 e3558f53
展开全部 显示空白变更内容
内联 并排
Showing with 0 addition and 145664 deletion
PaddleNLP/unarchived/chinese_ner/.run_ce.sh 已删除 100755 → 0
浏览文件 @ e3558f53
#!/bin/bash
export MKL_NUM_THREADS=1
export OMP_NUM_THREADS=1
cudaid=${chinese_ner:=0} # use 0-th card as default
export CUDA_VISIBLE_DEVICES=$cudaid
FLAGS_benchmark=true python train.py --num_passes 300 --device GPU --enable_ce | python _ce.py
cudaid=${chinese_ner_4:=0,1,2,3} # use 0-th card as default
export CUDA_VISIBLE_DEVICES=$cudaid
FLAGS_benchmark=true python train.py --num_passes 300 --device GPU --parallel True --enable_ce | python _ce.py
export CPU_NUM=1
export NUM_THREADS=1
FLAGS_benchmark=true python train.py --num_passes 300 --device CPU --enable_ce | python _ce.py
PaddleNLP/unarchived/chinese_ner/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
# 使用ParallelExecutor的中文命名实体识别示例
以下是本例的简要目录结构及说明：
```text
.
├── data # 存储运行本例所依赖的数据，从外部获取
├── reader.py # 数据读取接口, 从外部获取
├── README.md # 文档
├── train.py # 训练脚本
├── infer.py # 预测脚本
```
## 数据
在data目录下，有两个文件夹，train_files中保存的是训练数据，test_files中保存的是测试数据，作为示例，在目录下我们各放置了两个文件，实际训练时，根据自己的实际需要将数据放置在对应目录，并根据数据格式，修改reader.py中的数据读取函数。
## 训练
通过运行
```
python train.py --help
```
来获取命令行参数的帮助，设置正确的数据路径等参数后，运行`train.py`开始训练。
训练记录形如
```txt
pass_id:0, time_cost:4.92960214615s
[Train] precision:0.000862136531076, recall:0.0059880239521, f1:0.00150726226363
[Test] precision:0.000796178343949, recall:0.00335758254057, f1:0.00128713933283
pass_id:1, time_cost:0.715255975723s
[Train] precision:0.00474094141551, recall:0.00762112139358, f1:0.00584551148225
[Test] precision:0.0228873239437, recall:0.00727476217124, f1:0.0110403397028
pass_id:2, time_cost:0.740842103958s
[Train] precision:0.0120967741935, recall:0.00163309744148, f1:0.00287769784173
[Test] precision:0, recall:0.0, f1:0
```
## 预测
类似于训练过程，预测时指定需要测试模型的路径、测试数据、预测标记文件的路径，运行`infer.py`开始预测。
预测结果如下
```txt
152804 O O
130048 O O
38862 10-B O
784 O O
1540 O O
4145 O O
2255 O O
0 O O
1279 O O
7793 O O
373 O O
1621 O O
815 O O
2 O O
247 24-B O
401 24-I O
```
输出分为三列，以"\t"分割，第一列是输入的词语的序号，第二列是标准结果，第三列为标记结果。多条输入序列之间以空行分隔。
PaddleNLP/unarchived/chinese_ner/_ce.py 已删除 100644 → 0
浏览文件 @ e3558f53
# this file is only used for continuous evaluation test!
import os
import sys
sys.path.append(os.environ['ceroot'])
from kpi import CostKpi
from kpi import DurationKpi
from kpi import AccKpi
each_pass_duration_cpu1_thread1_kpi = DurationKpi(
'each_pass_duration_cpu1_thread1', 0.08, 0, actived=True)
train_recall_cpu1_thread1_kpi = AccKpi('train_recall_cpu1_thread1', 0.08, 0)
each_pass_duration_gpu1_kpi = DurationKpi(
'each_pass_duration_gpu1', 0.08, 0, actived=True)
train_recall_gpu1_kpi = AccKpi('train_recall_gpu1', 0.08, 0)
each_pass_duration_gpu4_kpi = DurationKpi(
'each_pass_duration_gpu4', 0.08, 0, actived=True)
train_recall_gpu4_kpi = AccKpi('train_recall_gpu4', 0.08, 0)
tracking_kpis = [
each_pass_duration_cpu1_thread1_kpi,
train_recall_cpu1_thread1_kpi,
each_pass_duration_gpu1_kpi,
train_recall_gpu1_kpi,
each_pass_duration_gpu4_kpi,
train_recall_gpu4_kpi,
]
def parse_log(log):
'''
This method should be implemented by model developers.
The suggestion:
each line in the log should be key, value, for example:
"
train_cost\t1.0
test_cost\t1.0
train_cost\t1.0
train_cost\t1.0
train_acc\t1.2
"
'''
for line in log.split('\n'):
fs = line.strip().split('\t')
print(fs)
if len(fs) == 3 and fs[0] == 'kpis':
kpi_name = fs[1]
kpi_value = float(fs[2])
yield kpi_name, kpi_value
def log_to_ce(log):
kpi_tracker = {}
for kpi in tracking_kpis:
kpi_tracker[kpi.name] = kpi
for (kpi_name, kpi_value) in parse_log(log):
print(kpi_name, kpi_value)
kpi_tracker[kpi_name].add_record(kpi_value)
kpi_tracker[kpi_name].persist()
if __name__ == '__main__':
log = sys.stdin.read()
log_to_ce(log)
PaddleNLP/unarchived/chinese_ner/data/label_dict 已删除 100644 → 0
浏览文件 @ e3558f53
24-B
24-I
27-B
27-I
20-B
20-I
21-B
21-I
22-B
22-I
23-B
23-I
28-B
28-I
29-B
29-I
12-B
12-I
11-B
11-I
10-B
10-I
13-B
13-I
38-B
38-I
14-B
14-I
16-B
16-I
33-B
33-I
18-B
18-I
31-B
31-I
30-B
30-I
37-B
37-I
36-B
36-I
35-B
35-I
19-B
19-I
32-B
32-I
O
PaddleNLP/unarchived/chinese_ner/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import numpy as np
import argparse
import time
import paddle.fluid as fluid
import paddle.fluid.profiler as profiler
import paddle
import reader
def parse_args():
parser = argparse.ArgumentParser("Run inference.")
parser.add_argument(
'--batch_size',
type=int,
default=6,
help='The size of a batch. (default: %(default)d)')
parser.add_argument(
'--device',
type=str,
default='GPU',
choices=['CPU', 'GPU'],
help='The device type. (default: %(default)s)')
parser.add_argument(
'--model_path',
type=str,
default='output/params_pass_0',
help='A path to the model. (default: %(default)s)')
parser.add_argument(
'--test_data_dir',
type=str,
default='data/test_files',
help='A directory with test data files. (default: %(default)s)')
parser.add_argument(
'--test_label_file',
type=str,
default='data/label_dict',
help='A file with test labels. (default: %(default)s)')
parser.add_argument(
'--num_passes', type=int, default=1, help='The number of passes.')
parser.add_argument(
'--skip_pass_num',
type=int,
default=0,
help='The first num of passes to skip in statistics calculations.')
parser.add_argument(
'--profile', action='store_true', help='If set, do profiling.')
args = parser.parse_args()
return args
def print_arguments(args):
print('----------- Configuration Arguments -----------')
for arg, value in sorted(vars(args).items()):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
def load_reverse_dict(dict_path):
return dict((idx, line.strip().split("\t")[0])
for idx, line in enumerate(open(dict_path, "r").readlines()))
def to_lodtensor(data, place):
seq_lens = [len(seq) for seq in data]
cur_len = 0
lod = [cur_len]
for l in seq_lens:
cur_len += l
lod.append(cur_len)
flattened_data = np.concatenate(data, axis=0).astype("int64")
flattened_data = flattened_data.reshape([len(flattened_data), 1])
res = fluid.LoDTensor()
res.set(flattened_data, place)
res.set_lod([lod])
return res
def infer(args):
word = fluid.layers.data(name='word', shape=[1], dtype='int64', lod_level=1)
mention = fluid.layers.data(
name='mention', shape=[1], dtype='int64', lod_level=1)
target = fluid.layers.data(
name='target', shape=[1], dtype='int64', lod_level=1)
label_reverse_dict = load_reverse_dict(args.test_label_file)
test_data = paddle.batch(
reader.file_reader(args.test_data_dir), batch_size=args.batch_size)
place = fluid.CUDAPlace(0) if args.device == 'GPU' else fluid.CPUPlace()
feeder = fluid.DataFeeder(feed_list=[word, mention, target], place=place)
exe = fluid.Executor(place)
inference_scope = fluid.Scope()
with fluid.scope_guard(inference_scope):
[inference_program, feed_target_names,
fetch_targets] = fluid.io.load_inference_model(args.model_path, exe)
total_passes = args.num_passes + args.skip_pass_num
batch_times = [0] * total_passes
word_counts = [0] * total_passes
wpses = [0] * total_passes
all_iters = 0
for pass_id in range(total_passes):
if pass_id < args.skip_pass_num:
print("Warm-up pass")
if pass_id == args.skip_pass_num:
profiler.reset_profiler()
iters = 0
for data in test_data():
word = to_lodtensor(list(map(lambda x: x[0], data)), place)
mention = to_lodtensor(list(map(lambda x: x[1], data)), place)
start = time.time()
crf_decode = exe.run(inference_program,
feed={"word": word,
"mention": mention},
fetch_list=fetch_targets,
return_numpy=False)
batch_time = time.time() - start
lod_info = (crf_decode[0].lod())[0]
np_data = np.array(crf_decode[0])
word_count = 0
assert len(data) == len(lod_info) - 1
for sen_index in range(len(data)):
assert len(data[sen_index][0]) == lod_info[
sen_index + 1] - lod_info[sen_index]
word_index = 0
for tag_index in range(lod_info[sen_index],
lod_info[sen_index + 1]):
word = str(data[sen_index][0][word_index])
gold_tag = label_reverse_dict[data[sen_index][2][
word_index]]
tag = label_reverse_dict[np_data[tag_index][0]]
word_index += 1
word_count += word_index
batch_times[pass_id] += batch_time
word_counts[pass_id] += word_count
iters += 1
all_iters += 1
batch_times[pass_id] /= iters
word_counts[pass_id] /= iters
wps = word_counts[pass_id] / batch_times[pass_id]
wpses[pass_id] = wps
print(
"Pass: %d, iterations (total): %d (%d), latency: %.5f s, words: %d, wps: %f"
% (pass_id, iters, all_iters, batch_times[pass_id],
word_counts[pass_id], wps))
# Postprocess benchmark data
latencies = batch_times[args.skip_pass_num:]
latency_avg = np.average(latencies)
latency_std = np.std(latencies)
latency_pc99 = np.percentile(latencies, 99)
wps_avg = np.average(wpses)
wps_std = np.std(wpses)
wps_pc01 = np.percentile(wpses, 1)
# Benchmark output
print('\nTotal passes (incl. warm-up): %d' % (total_passes))
print('Total iterations (incl. warm-up): %d' % (all_iters))
print('Total examples (incl. warm-up): %d' % (all_iters * args.batch_size))
print('avg latency: %.5f, std latency: %.5f, 99pc latency: %.5f' %
(latency_avg, latency_std, latency_pc99))
print('avg wps: %.5f, std wps: %.5f, wps for 99pc latency: %.5f' %
(wps_avg, wps_std, wps_pc01))
if __name__ == "__main__":
args = parse_args()
print_arguments(args)
if args.profile:
if args.device == 'GPU':
with profiler.cuda_profiler("cuda_profiler.txt", 'csv') as nvprof:
infer(args)
else:
with profiler.profiler('CPU', sorted_key='total') as cpuprof:
infer(args)
else:
infer(args)
PaddleNLP/unarchived/chinese_ner/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import os
def file_reader(file_dir):
def reader():
files = os.listdir(file_dir)
for fi in files:
for line in open(file_dir + '/' + fi, 'r'):
line = line.strip()
features = line.split(";")
word_idx = []
for item in features[1].strip().split(" "):
word_idx.append(int(item))
target_idx = []
for item in features[2].strip().split(" "):
label_index = int(item)
if label_index == 0:
label_index = 48
else:
label_index -= 1
target_idx.append(label_index)
mention_idx = []
for item in features[3].strip().split(" "):
mention_idx.append(int(item))
yield word_idx, mention_idx, target_idx,
return reader
PaddleNLP/unarchived/chinese_ner/scripts/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
## Purpose of this directory
The purpose of this directory is to provide exemplary execution commands. They are inside bash scripts described below.
## Preparation
To add execution permissions for shell scripts, run in this directory:
`chmod +x *.sh`
## Performance tips
Use the below environment flags for best performance:
```
KMP_AFFINITY=granularity=fine,compact,1,0
OMP_NUM_THREADS=<num_of_physical_cores>
```
For example, you can export them, or add them inside the specific files.
## Training
### CPU with mkldnn
Run:
`./train.sh MKLDNN`
### CPU without mkldnn
Run:
`./train.sh CPU`
### GPU
Run:
`./train.sh GPU`
## Inference
### CPU with mkldnn
Run:
`./infer.sh MKLDNN`
### CPU without mkldnn
Run:
`./infer.sh CPU`
### GPU
Run:
`./infer.sh GPU`
PaddleNLP/unarchived/chinese_ner/scripts/infer.sh 已删除 100644 → 0
浏览文件 @ e3558f53
#!/bin/bash
export MKL_NUM_THREADS=1
export OMP_NUM_THREADS=1
mode=$1 # gpu, cpu, mkldnn
if [ "$mode" = "CPU" ]; then
device="CPU"
model_path="cpu_model"
elif [ "$mode" = "GPU" ]; then
device="GPU"
model_path="gpu_model"
elif [ "$mode" = "MKLDNN" ]; then
device="CPU"
model_path="mkldnn_model"
export FLAGS_use_mkldnn=1
else
echo "Invalid mode provided. Please use one of {GPU, CPU, MKLDNN}"
exit 1
fi
ht=`lscpu |grep "per core"|awk -F':' '{print $2}'|xargs`
if [ $ht -eq 1 ]; then # HT is OFF
if [ -z "$KMP_AFFINITY" ]; then
export KMP_AFFINITY="granularity=fine,compact,0,0"
fi
if [ -z "$OMP_DYNAMIC" ]; then
export OMP_DYNAMIC="FALSE"
fi
else # HT is ON
if [ -z "$KMP_AFFINITY" ]; then
export KMP_AFFINITY="granularity=fine,compact,1,0"
fi
fi
python ../infer.py \
--device $device \
--num_passes 1 \
--skip_pass_num 2 \
--profile \
--test_data_dir ../data/test_files \
--test_label_file ../data/label_dict \
--model_path $model_path/params_pass_0
PaddleNLP/unarchived/chinese_ner/scripts/train.sh 已删除 100644 → 0
浏览文件 @ e3558f53
#!/bin/bash
export MKL_NUM_THREADS=1
export OMP_NUM_THREADS=1
mode=$1 # gpu, cpu, mkldnn
if [ "$mode" = "CPU" ]; then
device="CPU"
parallel="--parallel True"
save_model_dir="cpu_model"
elif [ "$mode" = "GPU" ]; then
device="GPU"
parallel="--parallel True"
save_model_dir="gpu_model"
elif [ "$mode" = "MKLDNN" ]; then
device="CPU"
parallel=""
save_model_dir="mkldnn_model"
export FLAGS_use_mkldnn=1
else
echo "Invalid mode provided. Please use one of {GPU, CPU, MKLDNN}"
exit 1
fi
ht=`lscpu |grep "per core"|awk -F':' '{print $2}'|xargs`
if [ $ht -eq 1 ]; then # HT is OFF
if [ -z "$KMP_AFFINITY" ]; then
export KMP_AFFINITY="granularity=fine,compact,0,0"
fi
if [ -z "$OMP_DYNAMIC" ]; then
export OMP_DYNAMIC="FALSE"
fi
else # HT is ON
if [ -z "$KMP_AFFINITY" ]; then
export KMP_AFFINITY="granularity=fine,compact,1,0"
fi
fi
python ../train.py \
--device $device \
$parallel \
--model_save_dir $save_model_dir \
--test_data_dir ../data/test_files \
--train_data_dir ../data/train_files \
--num_passes 1
PaddleNLP/unarchived/chinese_ner/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import os
import math
import time
import argparse
import numpy as np
import paddle
import paddle.fluid as fluid
from paddle.fluid.initializer import NormalInitializer
import reader
def parse_args():
parser = argparse.ArgumentParser("Run training.")
parser.add_argument(
'--batch_size',
type=int,
default=256,
help='The size of a batch. (default: %(default)d)')
parser.add_argument(
'--word_dict_len',
type=int,
default=1942563,
help='The lenght of the word dictionary. (default: %(default)d)')
parser.add_argument(
'--label_dict_len',
type=int,
default=49,
help='The lenght of the label dictionary. (default: %(default)d)')
parser.add_argument(
'--device',
type=str,
default='GPU',
choices=['CPU', 'GPU'],
help='The device type. (default: %(default)s)')
parser.add_argument(
'--train_data_dir',
type=str,
default='data/train_files',
help='A directory with train data files. (default: %(default)s)')
parser.add_argument(
'--parallel',
type=bool,
default=False,
help="Whether to use parallel training. (default: %(default)s)")
parser.add_argument(
'--test_data_dir',
type=str,
default='data/test_files',
help='A directory with test data files. (default: %(default)s)')
parser.add_argument(
'--model_save_dir',
type=str,
default='./output',
help='A directory for saving models. (default: %(default)s)')
parser.add_argument(
'--num_passes',
type=int,
default=1000,
help='The number of epochs. (default: %(default)d)')
parser.add_argument(
'--enable_ce',
action='store_true',
help='If set, run the task with continuous evaluation logs.')
args = parser.parse_args()
return args
def print_arguments(args):
print('----------- Configuration Arguments -----------')
for arg, value in sorted(vars(args).items()):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
def load_reverse_dict(dict_path):
return dict((idx, line.strip().split("\t")[0])
for idx, line in enumerate(open(dict_path, "r").readlines()))
def to_lodtensor(data, place):
seq_lens = [len(seq) for seq in data]
cur_len = 0
lod = [cur_len]
for l in seq_lens:
cur_len += l
lod.append(cur_len)
flattened_data = np.concatenate(data, axis=0).astype("int64")
flattened_data = flattened_data.reshape([len(flattened_data), 1])
res = fluid.LoDTensor()
res.set(flattened_data, place)
res.set_lod([lod])
return res
def ner_net(word_dict_len, label_dict_len):
IS_SPARSE = False
word_dim = 32
mention_dict_len = 57
mention_dim = 20
grnn_hidden = 36
emb_lr = 5
init_bound = 0.1
def _net_conf(word, mark, target):
word_embedding = fluid.layers.embedding(
input=word,
size=[word_dict_len, word_dim],
dtype='float32',
is_sparse=IS_SPARSE,
param_attr=fluid.ParamAttr(
learning_rate=emb_lr,
name="word_emb",
initializer=fluid.initializer.Uniform(
low=-init_bound, high=init_bound)))
mention_embedding = fluid.layers.embedding(
input=mention,
size=[mention_dict_len, mention_dim],
dtype='float32',
is_sparse=IS_SPARSE,
param_attr=fluid.ParamAttr(
learning_rate=emb_lr,
name="mention_emb",
initializer=fluid.initializer.Uniform(
low=-init_bound, high=init_bound)))
word_embedding_r = fluid.layers.embedding(
input=word,
size=[word_dict_len, word_dim],
dtype='float32',
is_sparse=IS_SPARSE,
param_attr=fluid.ParamAttr(
learning_rate=emb_lr,
name="word_emb_r",
initializer=fluid.initializer.Uniform(
low=-init_bound, high=init_bound)))
mention_embedding_r = fluid.layers.embedding(
input=mention,
size=[mention_dict_len, mention_dim],
dtype='float32',
is_sparse=IS_SPARSE,
param_attr=fluid.ParamAttr(
learning_rate=emb_lr,
name="mention_emb_r",
initializer=fluid.initializer.Uniform(
low=-init_bound, high=init_bound)))
word_mention_vector = fluid.layers.concat(
input=[word_embedding, mention_embedding], axis=1)
word_mention_vector_r = fluid.layers.concat(
input=[word_embedding_r, mention_embedding_r], axis=1)
pre_gru = fluid.layers.fc(
input=word_mention_vector,
size=grnn_hidden * 3,
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Uniform(
low=-init_bound, high=init_bound),
regularizer=fluid.regularizer.L2DecayRegularizer(
regularization_coeff=1e-4)))
gru = fluid.layers.dynamic_gru(
input=pre_gru,
size=grnn_hidden,
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Uniform(
low=-init_bound, high=init_bound),
regularizer=fluid.regularizer.L2DecayRegularizer(
regularization_coeff=1e-4)))
pre_gru_r = fluid.layers.fc(
input=word_mention_vector_r,
size=grnn_hidden * 3,
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Uniform(
low=-init_bound, high=init_bound),
regularizer=fluid.regularizer.L2DecayRegularizer(
regularization_coeff=1e-4)))
gru_r = fluid.layers.dynamic_gru(
input=pre_gru_r,
size=grnn_hidden,
is_reverse=True,
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Uniform(
low=-init_bound, high=init_bound),
regularizer=fluid.regularizer.L2DecayRegularizer(
regularization_coeff=1e-4)))
gru_merged = fluid.layers.concat(input=[gru, gru_r], axis=1)
emission = fluid.layers.fc(
size=label_dict_len,
input=gru_merged,
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Uniform(
low=-init_bound, high=init_bound),
regularizer=fluid.regularizer.L2DecayRegularizer(
regularization_coeff=1e-4)))
crf_cost = fluid.layers.linear_chain_crf(
input=emission,
label=target,
param_attr=fluid.ParamAttr(
name='crfw',
learning_rate=0.2, ))
avg_cost = fluid.layers.mean(x=crf_cost)
return avg_cost, emission
word = fluid.layers.data(name='word', shape=[1], dtype='int64', lod_level=1)
mention = fluid.layers.data(
name='mention', shape=[1], dtype='int64', lod_level=1)
target = fluid.layers.data(
name="target", shape=[1], dtype='int64', lod_level=1)
avg_cost, emission = _net_conf(word, mention, target)
return avg_cost, emission, word, mention, target
def test2(exe, chunk_evaluator, inference_program, test_data, place,
cur_fetch_list):
chunk_evaluator.reset()
for data in test_data():
word = to_lodtensor(list(map(lambda x: x[0], data)), place)
mention = to_lodtensor(list(map(lambda x: x[1], data)), place)
target = to_lodtensor(list(map(lambda x: x[2], data)), place)
result_list = exe.run(
inference_program,
feed={"word": word,
"mention": mention,
"target": target},
fetch_list=cur_fetch_list)
number_infer = np.array(result_list[0])
number_label = np.array(result_list[1])
number_correct = np.array(result_list[2])
chunk_evaluator.update(number_infer[0].astype('int64'),
number_label[0].astype('int64'),
number_correct[0].astype('int64'))
return chunk_evaluator.eval()
def test(test_exe, chunk_evaluator, inference_program, test_data, place,
cur_fetch_list):
chunk_evaluator.reset()
for data in test_data():
word = to_lodtensor(list(map(lambda x: x[0], data)), place)
mention = to_lodtensor(list(map(lambda x: x[1], data)), place)
target = to_lodtensor(list(map(lambda x: x[2], data)), place)
result_list = test_exe.run(
fetch_list=cur_fetch_list,
feed={"word": word,
"mention": mention,
"target": target})
number_infer = np.array(result_list[0])
number_label = np.array(result_list[1])
number_correct = np.array(result_list[2])
chunk_evaluator.update(number_infer.sum().astype('int64'),
number_label.sum().astype('int64'),
number_correct.sum().astype('int64'))
return chunk_evaluator.eval()
def main(args):
if not os.path.exists(args.model_save_dir):
os.makedirs(args.model_save_dir)
main = fluid.Program()
startup = fluid.Program()
if args.enable_ce:
SEED = 102
main.random_seed = SEED
startup.random_seed = SEED
with fluid.program_guard(main, startup):
avg_cost, feature_out, word, mention, target = ner_net(
args.word_dict_len, args.label_dict_len)
crf_decode = fluid.layers.crf_decoding(
input=feature_out, param_attr=fluid.ParamAttr(name='crfw'))
(precision, recall, f1_score, num_infer_chunks, num_label_chunks,
num_correct_chunks) = fluid.layers.chunk_eval(
input=crf_decode,
label=target,
chunk_scheme="IOB",
num_chunk_types=int(math.ceil((args.label_dict_len - 1) / 2.0)))
inference_program = fluid.default_main_program().clone(for_test=True)
sgd_optimizer = fluid.optimizer.SGD(learning_rate=1e-3)
sgd_optimizer.minimize(avg_cost)
chunk_evaluator = fluid.metrics.ChunkEvaluator()
train_reader = paddle.batch(
paddle.reader.shuffle(
reader.file_reader(args.train_data_dir), buf_size=2000000),
batch_size=args.batch_size)
test_reader = paddle.batch(
paddle.reader.shuffle(
reader.file_reader(args.test_data_dir), buf_size=2000000),
batch_size=args.batch_size)
place = fluid.CUDAPlace(0) if args.device == 'GPU' else fluid.CPUPlace()
feeder = fluid.DataFeeder(
feed_list=[word, mention, target], place=place)
exe = fluid.Executor(place)
exe.run(startup)
if args.parallel:
train_exe = fluid.ParallelExecutor(
loss_name=avg_cost.name, use_cuda=(args.device == 'GPU'))
test_exe = fluid.ParallelExecutor(
use_cuda=(args.device == 'GPU'),
main_program=inference_program,
share_vars_from=train_exe)
else:
train_exe = exe
test_exe = exe
total_time = 0
ce_info = []
batch_id = 0
for pass_id in range(args.num_passes):
chunk_evaluator.reset()
train_reader_iter = train_reader()
start_time = time.time()
while True:
try:
cur_batch = next(train_reader_iter)
cost, nums_infer, nums_label, nums_correct = train_exe.run(
fetch_list=[
avg_cost.name, num_infer_chunks.name,
num_label_chunks.name, num_correct_chunks.name
],
feed=feeder.feed(cur_batch))
chunk_evaluator.update(
np.array(nums_infer).sum().astype("int64"),
np.array(nums_label).sum().astype("int64"),
np.array(nums_correct).sum().astype("int64"))
cost_list = np.array(cost)
batch_id += 1
except StopIteration:
break
end_time = time.time()
total_time += end_time - start_time
print("pass_id:" + str(pass_id) + ", time_cost:" + str(
end_time - start_time) + "s")
precision, recall, f1_score = chunk_evaluator.eval()
print("[Train] precision:" + str(precision) + ", recall:" + str(
recall) + ", f1:" + str(f1_score))
ce_info.append(recall)
p, r, f1 = test2(
exe, chunk_evaluator, inference_program, test_reader, place,
[num_infer_chunks, num_label_chunks, num_correct_chunks])
print("[Test] precision:" + str(p) + ", recall:" + str(r) + ", f1:"
+ str(f1))
save_dirname = os.path.join(args.model_save_dir,
"params_pass_%d" % pass_id)
fluid.io.save_inference_model(save_dirname, ['word', 'mention'],
[crf_decode], exe)
# only for ce
if args.enable_ce:
ce_recall = 0
try:
ce_recall = ce_info[-2]
except:
print("ce info error")
epoch_idx = args.num_passes
device = get_device(args)
if args.device == "GPU":
gpu_num = device[1]
print("kpis\teach_pass_duration_gpu%s\t%s" %
(gpu_num, total_time / epoch_idx))
print("kpis\ttrain_recall_gpu%s\t%s" % (gpu_num, ce_recall))
else:
cpu_num = device[1]
threads_num = device[2]
print("kpis\teach_pass_duration_cpu%s_thread%s\t%s" %
(cpu_num, threads_num, total_time / epoch_idx))
print("kpis\ttrain_recall_cpu%s_thread%s\t%s" %
(cpu_num, threads_num, ce_recall))
def get_device(args):
if args.device == "GPU":
gpus = os.environ.get("CUDA_VISIBLE_DEVICES", "")
gpu_num = len(gpus.split(','))
return "gpu", gpu_num
else:
threads_num = os.environ.get('NUM_THREADS', 1)
cpu_num = os.environ.get('CPU_NUM', 1)
return "cpu", int(cpu_num), int(threads_num)
if __name__ == "__main__":
args = parse_args()
print_arguments(args)
main(args)
PaddleNLP/unarchived/deep_attention_matching_net/.run_ce.sh 已删除 100755 → 0
浏览文件 @ e3558f53
###!/bin/bash
####This file is only used for continuous evaluation.
export CE_MODE_X=1
export CUDA_VISIBLE_DEVICES=0
export FLAGS_eager_delete_tensor_gb=0.0
if [ ! -e data_small.pkl ]; then
wget -c http://dam-data.bj.bcebos.com/data_small.pkl
fi
python train_and_evaluate.py --data_path data_small.pkl \
--use_cuda \
--use_pyreader \
--num_scan_data 1 \
--batch_size 100 | python _ce.py
PaddleNLP/unarchived/deep_attention_matching_net/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
# __Deep Attention Matching Network__
This is the source code of Deep Attention Matching network (DAM), that is proposed for multi-turn response selection in the retrieval-based chatbot.
DAM is a neural matching network that entirely based on attention mechanism. The motivation of DAM is to capture those semantic dependencies, among dialogue elements at different level of granularities, in multi-turn conversation as matching evidences, in order to better match response candidate with its multi-turn context. DAM appears on ACL-2018, please find our paper at [http://aclweb.org/anthology/P18-1103](http://aclweb.org/anthology/P18-1103).
## __Network__
DAM is inspired by Transformer in Machine Translation (Vaswani et al., 2017), and we extend the key attention mechanism of Transformer in two perspectives and introduce those two kinds of attention in one uniform neural network.
- **self-attention** To gradually capture semantic representations in different granularities by stacking attention from word-level embeddings. Those multi-grained semantic representations would facilitate exploring segmental dependencies between context and response.
- **cross-attention** Attention across context and response can generally capture the relevance in dependency between segment pairs, which could provide complementary information to textual relevance for matching response with multi-turn context.
<p align="center">
<img src="images/Figure1.png"/> <br />
Overview of Deep Attention Matching Network
</p>
## __Results__
We test DAM on two large-scale multi-turn response selection tasks, i.e., the Ubuntu Corpus v1 and Douban Conversation Corpus, experimental results are bellow:
<p align="center">
<img src="images/Figure2.png"/> <br />
</p>
## __Usage__
Take the experiment on the Ubuntu Corpus v1 for Example.
1) Go to the `ubuntu` directory
```
cd ubuntu
```
2) Download the well-preprocessed data for training
```
sh download_data.sh
```
3) Execute the model training and evaluation by
```
sh train.sh
```
for more detailed explanation about the arguments, please run
```
python ../train_and_evaluate.py --help
```
By default, the training is executed on one single GPU, which can be switched to multiple-GPU mode easily by simply resetting the visible devices in `train.sh`, e.g.,
```
export CUDA_VISIBLE_DEVICES=0,1,2,3
```
4) Run test by
```
sh test.sh
```
and run the test for different saved models by using different argument `--model_path`.
Similary, one can carry out the experiment on the Douban Conversation Corpus by going to the directory `douban` and following the same procedure.
## __Dependencies__
- Python >= 2.7.3
- PaddlePaddle latest develop branch
## __Citation__
The following article describe the DAM in detail. We recommend citing this article as default.
```
@inproceedings{ ,
title={Multi-Turn Response Selection for Chatbots with Deep Attention Matching Network},
author={Xiangyang Zhou, Lu Li, Daxiang Dong, Yi Liu, Ying Chen, Wayne Xin Zhao, Dianhai Yu and Hua Wu},
booktitle={Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
volume={1},
pages={ -- },
year={2018}
}
```
PaddleNLP/unarchived/deep_attention_matching_net/_ce.py 已删除 100644 → 0
浏览文件 @ e3558f53
####this file is only used for continuous evaluation test!
import os
import sys
sys.path.append(os.environ['ceroot'])
from kpi import CostKpi, DurationKpi, AccKpi
#### NOTE kpi.py should shared in models in some way!!!!
train_cost_kpi = CostKpi('train_cost', 0.02, 0, actived=True)
train_duration_kpi = DurationKpi('train_duration', 0.05, 0, actived=True)
tracking_kpis = [
train_cost_kpi,
train_duration_kpi,
]
def parse_log(log):
for line in log.split('\n'):
fs = line.strip().split('\t')
print(fs)
if len(fs) == 3 and fs[0] == 'kpis':
print("-----%s" % fs)
kpi_name = fs[1]
kpi_value = float(fs[2])
yield kpi_name, kpi_value
def log_to_ce(log):
kpi_tracker = {}
for kpi in tracking_kpis:
kpi_tracker[kpi.name] = kpi
for (kpi_name, kpi_value) in parse_log(log):
print(kpi_name, kpi_value)
kpi_tracker[kpi_name].add_record(kpi_value)
kpi_tracker[kpi_name].persist()
if __name__ == '__main__':
log = sys.stdin.read()
print("*****")
print(log)
print("****")
log_to_ce(log)
PaddleNLP/unarchived/deep_attention_matching_net/douban/download_data.sh 已删除 100644 → 0
浏览文件 @ e3558f53
url=http://dam-data.cdn.bcebos.com/douban.tar.gz
md5=e07ca68f21c20e09efb3e8b247194405
if [ ! -e douban.tar.gz ]; then
wget -c $url
fi
echo "Checking md5 sum ..."
md5sum_tmp=`md5sum douban.tar.gz | cut -d ' ' -f1`
if [ $md5sum_tmp != $md5 ]; then
echo "Md5sum check failed, please remove and redownload douban.tar.gz"
exit 1
fi
echo "Untar douban.tar.gz ..."
tar -xzvf douban.tar.gz
PaddleNLP/unarchived/deep_attention_matching_net/douban/test.sh 已删除 100644 → 0
浏览文件 @ e3558f53
export CUDA_VISIBLE_DEVICES=0
python -u ../test_and_evaluate.py --use_cuda \
--ext_eval \
--data_path ./data/data.pkl \
--save_path ./eval_3900 \
--model_path models/step_3900 \
--channel1_num 16 \
--batch_size 200 \
--vocab_size 172130 \
--emb_size 200 \
--_EOS_ 1
PaddleNLP/unarchived/deep_attention_matching_net/douban/train.sh 已删除 100644 → 0
浏览文件 @ e3558f53
export CUDA_VISIBLE_DEVICES=0
export FLAGS_eager_delete_tensor_gb=0.0
python -u ../train_and_evaluate.py --use_cuda \
--data_path ./data/data.pkl \
--ext_eval \
--word_emb_init ./data/word_embedding.pkl \
--save_path ./models \
--use_pyreader \
--batch_size 256 \
--vocab_size 172130 \
--channel1_num 16 \
--emb_size 200 \
--_EOS_ 1
PaddleNLP/unarchived/deep_attention_matching_net/model.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import six
import numpy as np
import paddle.fluid as fluid
import utils.layers as layers
class Net(object):
def __init__(self, max_turn_num, max_turn_len, vocab_size, emb_size,
stack_num, channel1_num, channel2_num):
self._max_turn_num = max_turn_num
self._max_turn_len = max_turn_len
self._vocab_size = vocab_size
self._emb_size = emb_size
self._stack_num = stack_num
self._channel1_num = channel1_num
self._channel2_num = channel2_num
self._feed_names = []
self.word_emb_name = "shared_word_emb"
self.use_stack_op = True
self.use_mask_cache = True
self.use_sparse_embedding = True
def create_py_reader(self, capacity, name):
# turns ids
shapes = [[-1, self._max_turn_len, 1]
for i in six.moves.xrange(self._max_turn_num)]
dtypes = ["int64" for i in six.moves.xrange(self._max_turn_num)]
# turns mask
shapes += [[-1, self._max_turn_len, 1]
for i in six.moves.xrange(self._max_turn_num)]
dtypes += ["float32" for i in six.moves.xrange(self._max_turn_num)]
# response ids, response mask, label
shapes += [[-1, self._max_turn_len, 1], [-1, self._max_turn_len, 1],
[-1, 1]]
dtypes += ["int64", "float32", "float32"]
py_reader = fluid.layers.py_reader(
capacity=capacity,
shapes=shapes,
lod_levels=[0] * (2 * self._max_turn_num + 3),
dtypes=dtypes,
name=name,
use_double_buffer=True)
data_vars = fluid.layers.read_file(py_reader)
self.turns_data = data_vars[0:self._max_turn_num]
self.turns_mask = data_vars[self._max_turn_num:2 * self._max_turn_num]
self.response = data_vars[-3]
self.response_mask = data_vars[-2]
self.label = data_vars[-1]
return py_reader
def create_data_layers(self):
self._feed_names = []
self.turns_data = []
for i in six.moves.xrange(self._max_turn_num):
name = "turn_%d" % i
turn = fluid.layers.data(
name=name, shape=[self._max_turn_len, 1], dtype="int64")
self.turns_data.append(turn)
self._feed_names.append(name)
self.turns_mask = []
for i in six.moves.xrange(self._max_turn_num):
name = "turn_mask_%d" % i
turn_mask = fluid.layers.data(
name=name, shape=[self._max_turn_len, 1], dtype="float32")
self.turns_mask.append(turn_mask)
self._feed_names.append(name)
self.response = fluid.layers.data(
name="response", shape=[self._max_turn_len, 1], dtype="int64")
self.response_mask = fluid.layers.data(
name="response_mask",
shape=[self._max_turn_len, 1],
dtype="float32")
self.label = fluid.layers.data(name="label", shape=[1], dtype="float32")
self._feed_names += ["response", "response_mask", "label"]
def get_feed_names(self):
return self._feed_names
def set_word_embedding(self, word_emb, place):
word_emb_param = fluid.global_scope().find_var(
self.word_emb_name).get_tensor()
word_emb_param.set(word_emb, place)
def create_network(self):
mask_cache = dict() if self.use_mask_cache else None
response_emb = fluid.layers.embedding(
input=self.response,
size=[self._vocab_size + 1, self._emb_size],
is_sparse=self.use_sparse_embedding,
param_attr=fluid.ParamAttr(
name=self.word_emb_name,
initializer=fluid.initializer.Normal(scale=0.1)))
# response part
Hr = response_emb
Hr_stack = [Hr]
for index in six.moves.xrange(self._stack_num):
Hr = layers.block(
name="response_self_stack" + str(index),
query=Hr,
key=Hr,
value=Hr,
d_key=self._emb_size,
q_mask=self.response_mask,
k_mask=self.response_mask,
mask_cache=mask_cache)
Hr_stack.append(Hr)
# context part
sim_turns = []
for t in six.moves.xrange(self._max_turn_num):
Hu = fluid.layers.embedding(
input=self.turns_data[t],
size=[self._vocab_size + 1, self._emb_size],
is_sparse=self.use_sparse_embedding,
param_attr=fluid.ParamAttr(
name=self.word_emb_name,
initializer=fluid.initializer.Normal(scale=0.1)))
Hu_stack = [Hu]
for index in six.moves.xrange(self._stack_num):
# share parameters
Hu = layers.block(
name="turn_self_stack" + str(index),
query=Hu,
key=Hu,
value=Hu,
d_key=self._emb_size,
q_mask=self.turns_mask[t],
k_mask=self.turns_mask[t],
mask_cache=mask_cache)
Hu_stack.append(Hu)
# cross attention
r_a_t_stack = []
t_a_r_stack = []
for index in six.moves.xrange(self._stack_num + 1):
t_a_r = layers.block(
name="t_attend_r_" + str(index),
query=Hu_stack[index],
key=Hr_stack[index],
value=Hr_stack[index],
d_key=self._emb_size,
q_mask=self.turns_mask[t],
k_mask=self.response_mask,
mask_cache=mask_cache)
r_a_t = layers.block(
name="r_attend_t_" + str(index),
query=Hr_stack[index],
key=Hu_stack[index],
value=Hu_stack[index],
d_key=self._emb_size,
q_mask=self.response_mask,
k_mask=self.turns_mask[t],
mask_cache=mask_cache)
t_a_r_stack.append(t_a_r)
r_a_t_stack.append(r_a_t)
t_a_r_stack.extend(Hu_stack)
r_a_t_stack.extend(Hr_stack)
if self.use_stack_op:
t_a_r = fluid.layers.stack(t_a_r_stack, axis=1)
r_a_t = fluid.layers.stack(r_a_t_stack, axis=1)
else:
for index in six.moves.xrange(len(t_a_r_stack)):
t_a_r_stack[index] = fluid.layers.unsqueeze(
input=t_a_r_stack[index], axes=[1])
r_a_t_stack[index] = fluid.layers.unsqueeze(
input=r_a_t_stack[index], axes=[1])
t_a_r = fluid.layers.concat(input=t_a_r_stack, axis=1)
r_a_t = fluid.layers.concat(input=r_a_t_stack, axis=1)
# sim shape: [batch_size, 2*(stack_num+1), max_turn_len, max_turn_len]
sim = fluid.layers.matmul(
x=t_a_r, y=r_a_t, transpose_y=True, alpha=1 / np.sqrt(200.0))
sim_turns.append(sim)
if self.use_stack_op:
sim = fluid.layers.stack(sim_turns, axis=2)
else:
for index in six.moves.xrange(len(sim_turns)):
sim_turns[index] = fluid.layers.unsqueeze(
input=sim_turns[index], axes=[2])
# sim shape: [batch_size, 2*(stack_num+1), max_turn_num, max_turn_len, max_turn_len]
sim = fluid.layers.concat(input=sim_turns, axis=2)
final_info = layers.cnn_3d(sim, self._channel1_num, self._channel2_num)
loss, logits = layers.loss(final_info, self.label)
return loss, logits
PaddleNLP/unarchived/deep_attention_matching_net/test_and_evaluate.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import os
import six
import numpy as np
import time
import argparse
import multiprocessing
import paddle
import paddle.fluid as fluid
import utils.reader as reader
from utils.util import print_arguments, mkdir
try:
import cPickle as pickle #python 2
except ImportError as e:
import pickle #python 3
from model import Net
#yapf: disable
def parse_args():
parser = argparse.ArgumentParser("Test for DAM.")
parser.add_argument(
'--batch_size',
type=int,
default=256,
help='Batch size for training. (default: %(default)d)')
parser.add_argument(
'--num_scan_data',
type=int,
default=2,
help='Number of pass for training. (default: %(default)d)')
parser.add_argument(
'--learning_rate',
type=float,
default=1e-3,
help='Learning rate used to train. (default: %(default)f)')
parser.add_argument(
'--data_path',
type=str,
default="data/ubuntu/data_small.pkl",
help='Path to training data. (default: %(default)s)')
parser.add_argument(
'--save_path',
type=str,
default="./",
help='Path to save score and result files. (default: %(default)s)')
parser.add_argument(
'--model_path',
type=str,
default="saved_models/step_1000",
help='Path to load well-trained models. (default: %(default)s)')
parser.add_argument(
'--use_cuda',
action='store_true',
help='If set, use cuda for training.')
parser.add_argument(
'--ext_eval',
action='store_true',
help='If set, use MAP, MRR ect for evaluation.')
parser.add_argument(
'--max_turn_num',
type=int,
default=9,
help='Maximum number of utterances in context.')
parser.add_argument(
'--max_turn_len',
type=int,
default=50,
help='Maximum length of setences in turns.')
parser.add_argument(
'--word_emb_init',
type=str,
default=None,
help='Path to the initial word embedding.')
parser.add_argument(
'--vocab_size',
type=int,
default=434512,
help='The size of vocabulary.')
parser.add_argument(
'--emb_size',
type=int,
default=200,
help='The dimension of word embedding.')
parser.add_argument(
'--_EOS_',
type=int,
default=28270,
help='The id for end of sentence in vocabulary.')
parser.add_argument(
'--stack_num',
type=int,
default=5,
help='The number of stacked attentive modules in network.')
parser.add_argument(
'--channel1_num',
type=int,
default=32,
help="The channels' number of the 1st conv3d layer's output.")
parser.add_argument(
'--channel2_num',
type=int,
default=16,
help="The channels' number of the 2nd conv3d layer's output.")
args = parser.parse_args()
return args
#yapf: enable
def test(args):
if not os.path.exists(args.save_path):
mkdir(args.save_path)
if not os.path.exists(args.model_path):
raise ValueError("Invalid model init path %s" % args.model_path)
# data data_config
data_conf = {
"batch_size": args.batch_size,
"max_turn_num": args.max_turn_num,
"max_turn_len": args.max_turn_len,
"_EOS_": args._EOS_,
}
dam = Net(args.max_turn_num, args.max_turn_len, args.vocab_size,
args.emb_size, args.stack_num, args.channel1_num,
args.channel2_num)
dam.create_data_layers()
loss, logits = dam.create_network()
loss.persistable = True
logits.persistable = True
# gradient clipping
fluid.clip.set_gradient_clip(clip=fluid.clip.GradientClipByValue(
max=1.0, min=-1.0))
test_program = fluid.default_main_program().clone(for_test=True)
optimizer = fluid.optimizer.Adam(
learning_rate=fluid.layers.exponential_decay(
learning_rate=args.learning_rate,
decay_steps=400,
decay_rate=0.9,
staircase=True))
optimizer.minimize(loss)
if args.use_cuda:
place = fluid.CUDAPlace(0)
dev_count = fluid.core.get_cuda_device_count()
else:
place = fluid.CPUPlace()
dev_count = multiprocessing.cpu_count()
exe = fluid.Executor(place)
exe.run(fluid.default_startup_program())
fluid.io.load_persistables(exe, args.model_path)
test_exe = fluid.ParallelExecutor(
use_cuda=args.use_cuda, main_program=test_program)
print("start loading data ...")
with open(args.data_path, 'rb') as f:
if six.PY2:
train_data, val_data, test_data = pickle.load(f)
else:
train_data, val_data, test_data = pickle.load(f, encoding="bytes")
print("finish loading data ...")
if args.ext_eval:
import utils.douban_evaluation as eva
eval_metrics = ["MAP", "MRR", "P@1", "R_{10}@1", "R_{10}@2", "R_{10}@5"]
else:
import utils.evaluation as eva
eval_metrics = ["R_2@1", "R_{10}@1", "R_{10}@2", "R_{10}@5"]
test_batches = reader.build_batches(test_data, data_conf)
test_batch_num = len(test_batches["response"])
print("test batch num: %d" % test_batch_num)
print("begin inference ...")
print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))
score_path = os.path.join(args.save_path, 'score.txt')
score_file = open(score_path, 'w')
for it in six.moves.xrange(test_batch_num // dev_count):
feed_list = []
for dev in six.moves.xrange(dev_count):
index = it * dev_count + dev
batch_data = reader.make_one_batch_input(test_batches, index)
feed_dict = dict(zip(dam.get_feed_names(), batch_data))
feed_list.append(feed_dict)
predicts = test_exe.run(feed=feed_list, fetch_list=[logits.name])
scores = np.array(predicts[0])
print("step = %d" % it)
for dev in six.moves.xrange(dev_count):
index = it * dev_count + dev
for i in six.moves.xrange(args.batch_size):
score_file.write(
str(scores[args.batch_size * dev + i][0]) + '\t' + str(
test_batches["label"][index][i]) + '\n')
score_file.close()
#write evaluation result
result = eva.evaluate(score_path)
result_file_path = os.path.join(args.save_path, 'result.txt')
with open(result_file_path, 'w') as out_file:
for metric, p_at in zip(eval_metrics, result):
out_file.write(metric + ": " + str(p_at) + '\n')
print('finish test')
print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))
if __name__ == '__main__':
args = parse_args()
print_arguments(args)
test(args)
PaddleNLP/unarchived/deep_attention_matching_net/train_and_evaluate.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import os
import six
import numpy as np
import time
import argparse
import multiprocessing
import paddle
import paddle.fluid as fluid
import utils.reader as reader
from utils.util import print_arguments, mkdir
try:
import cPickle as pickle #python 2
except ImportError as e:
import pickle #python 3
from model import Net
#yapf: disable
def parse_args():
parser = argparse.ArgumentParser("Training DAM.")
parser.add_argument(
'--batch_size',
type=int,
default=256,
help='Batch size for training. (default: %(default)d)')
parser.add_argument(
'--num_scan_data',
type=int,
default=2,
help='Number of pass for training. (default: %(default)d)')
parser.add_argument(
'--learning_rate',
type=float,
default=1e-3,
help='Learning rate used to train. (default: %(default)f)')
parser.add_argument(
'--data_path',
type=str,
default="data/data_small.pkl",
help='Path to training data. (default: %(default)s)')
parser.add_argument(
'--save_path',
type=str,
default="saved_models",
help='Path to save trained models. (default: %(default)s)')
parser.add_argument(
'--use_cuda',
action='store_true',
help='If set, use cuda for training.')
parser.add_argument(
'--use_pyreader',
action='store_true',
help='If set, use pyreader for reading data.')
parser.add_argument(
'--ext_eval',
action='store_true',
help='If set, use MAP, MRR ect for evaluation.')
parser.add_argument(
'--max_turn_num',
type=int,
default=9,
help='Maximum number of utterances in context.')
parser.add_argument(
'--max_turn_len',
type=int,
default=50,
help='Maximum length of setences in turns.')
parser.add_argument(
'--word_emb_init',
type=str,
default=None,
help='Path to the initial word embedding.')
parser.add_argument(
'--vocab_size',
type=int,
default=434512,
help='The size of vocabulary.')
parser.add_argument(
'--emb_size',
type=int,
default=200,
help='The dimension of word embedding.')
parser.add_argument(
'--_EOS_',
type=int,
default=28270,
help='The id for the end of sentence in vocabulary.')
parser.add_argument(
'--stack_num',
type=int,
default=5,
help='The number of stacked attentive modules in network.')
parser.add_argument(
'--channel1_num',
type=int,
default=32,
help="The channels' number of the 1st conv3d layer's output.")
parser.add_argument(
'--channel2_num',
type=int,
default=16,
help="The channels' number of the 2nd conv3d layer's output.")
args = parser.parse_args()
return args
#yapf: enable
def evaluate(score_path, result_file_path):
if args.ext_eval:
import utils.douban_evaluation as eva
else:
import utils.evaluation as eva
#write evaluation result
result = eva.evaluate(score_path)
with open(result_file_path, 'w') as out_file:
for p_at in result:
out_file.write(str(p_at) + '\n')
print('finish evaluation')
print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))
def test_with_feed(exe, program, feed_names, fetch_list, score_path, batches,
batch_num, dev_count):
score_file = open(score_path, 'w')
for it in six.moves.xrange(batch_num // dev_count):
feed_list = []
for dev in six.moves.xrange(dev_count):
val_index = it * dev_count + dev
batch_data = reader.make_one_batch_input(batches, val_index)
feed_dict = dict(zip(feed_names, batch_data))
feed_list.append(feed_dict)
predicts = exe.run(feed=feed_list, fetch_list=fetch_list)
scores = np.array(predicts[0])
for dev in six.moves.xrange(dev_count):
val_index = it * dev_count + dev
for i in six.moves.xrange(args.batch_size):
score_file.write(
str(scores[args.batch_size * dev + i][0]) + '\t' + str(
batches["label"][val_index][i]) + '\n')
score_file.close()
def test_with_pyreader(exe, program, pyreader, fetch_list, score_path, batches,
batch_num, dev_count):
def data_provider():
for index in six.moves.xrange(batch_num):
yield reader.make_one_batch_input(batches, index)
score_file = open(score_path, 'w')
pyreader.decorate_tensor_provider(data_provider)
it = 0
pyreader.start()
while True:
try:
predicts = exe.run(fetch_list=fetch_list)
scores = np.array(predicts[0])
for dev in six.moves.xrange(dev_count):
val_index = it * dev_count + dev
for i in six.moves.xrange(args.batch_size):
score_file.write(
str(scores[args.batch_size * dev + i][0]) + '\t' + str(
batches["label"][val_index][i]) + '\n')
it += 1
except fluid.core.EOFException:
pyreader.reset()
break
score_file.close()
def train(args):
if not os.path.exists(args.save_path):
os.makedirs(args.save_path)
# data data_config
data_conf = {
"batch_size": args.batch_size,
"max_turn_num": args.max_turn_num,
"max_turn_len": args.max_turn_len,
"_EOS_": args._EOS_,
}
dam = Net(args.max_turn_num, args.max_turn_len, args.vocab_size,
args.emb_size, args.stack_num, args.channel1_num,
args.channel2_num)
train_program = fluid.Program()
train_startup = fluid.Program()
if "CE_MODE_X" in os.environ:
train_program.random_seed = 110
train_startup.random_seed = 110
with fluid.program_guard(train_program, train_startup):
with fluid.unique_name.guard():
if args.use_pyreader:
train_pyreader = dam.create_py_reader(
capacity=10, name='train_reader')
else:
dam.create_data_layers()
loss, logits = dam.create_network()
loss.persistable = True
logits.persistable = True
# gradient clipping
fluid.clip.set_gradient_clip(clip=fluid.clip.GradientClipByValue(
max=1.0, min=-1.0))
optimizer = fluid.optimizer.Adam(
learning_rate=fluid.layers.exponential_decay(
learning_rate=args.learning_rate,
decay_steps=400,
decay_rate=0.9,
staircase=True))
optimizer.minimize(loss)
test_program = fluid.Program()
test_startup = fluid.Program()
if "CE_MODE_X" in os.environ:
test_program.random_seed = 110
test_startup.random_seed = 110
with fluid.program_guard(test_program, test_startup):
with fluid.unique_name.guard():
if args.use_pyreader:
test_pyreader = dam.create_py_reader(
capacity=10, name='test_reader')
else:
dam.create_data_layers()
loss, logits = dam.create_network()
loss.persistable = True
logits.persistable = True
test_program = test_program.clone(for_test=True)
if args.use_cuda:
place = fluid.CUDAPlace(0)
dev_count = fluid.core.get_cuda_device_count()
else:
place = fluid.CPUPlace()
dev_count = int(os.environ.get('CPU_NUM', multiprocessing.cpu_count()))
print("device count %d" % dev_count)
print("theoretical memory usage: ")
print(fluid.contrib.memory_usage(
program=train_program, batch_size=args.batch_size))
exe = fluid.Executor(place)
exe.run(train_startup)
exe.run(test_startup)
train_exe = fluid.ParallelExecutor(
use_cuda=args.use_cuda, loss_name=loss.name, main_program=train_program)
test_exe = fluid.ParallelExecutor(
use_cuda=args.use_cuda,
main_program=test_program,
share_vars_from=train_exe)
if args.word_emb_init is not None:
print("start loading word embedding init ...")
if six.PY2:
word_emb = np.array(pickle.load(open(args.word_emb_init,
'rb'))).astype('float32')
else:
word_emb = np.array(
pickle.load(
open(args.word_emb_init, 'rb'), encoding="bytes")).astype(
'float32')
dam.set_word_embedding(word_emb, place)
print("finish init word embedding ...")
print("start loading data ...")
with open(args.data_path, 'rb') as f:
if six.PY2:
train_data, val_data, test_data = pickle.load(f)
else:
train_data, val_data, test_data = pickle.load(f, encoding="bytes")
print("finish loading data ...")
val_batches = reader.build_batches(val_data, data_conf)
batch_num = len(train_data[six.b('y')]) // args.batch_size
val_batch_num = len(val_batches["response"])
print_step = max(1, batch_num // (dev_count * 100))
save_step = max(1, batch_num // (dev_count * 10))
print("begin model training ...")
print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))
# train on one epoch data by feeding
def train_with_feed(step):
ave_cost = 0.0
for it in six.moves.xrange(batch_num // dev_count):
feed_list = []
for dev in six.moves.xrange(dev_count):
index = it * dev_count + dev
batch_data = reader.make_one_batch_input(train_batches, index)
feed_dict = dict(zip(dam.get_feed_names(), batch_data))
feed_list.append(feed_dict)
cost = train_exe.run(feed=feed_list, fetch_list=[loss.name])
ave_cost += np.array(cost[0]).mean()
step = step + 1
if step % print_step == 0:
print("processed: [" + str(step * dev_count * 1.0 / batch_num) +
"] ave loss: [" + str(ave_cost / print_step) + "]")
ave_cost = 0.0
if (args.save_path is not None) and (step % save_step == 0):
save_path = os.path.join(args.save_path, "step_" + str(step))
print("Save model at step %d ... " % step)
print(time.strftime('%Y-%m-%d %H:%M:%S',
time.localtime(time.time())))
fluid.io.save_persistables(exe, save_path, train_program)
score_path = os.path.join(args.save_path, 'score.' + str(step))
test_with_feed(test_exe, test_program,
dam.get_feed_names(), [logits.name], score_path,
val_batches, val_batch_num, dev_count)
result_file_path = os.path.join(args.save_path,
'result.' + str(step))
evaluate(score_path, result_file_path)
return step, np.array(cost[0]).mean()
# train on one epoch with pyreader
def train_with_pyreader(step):
def data_provider():
for index in six.moves.xrange(batch_num):
yield reader.make_one_batch_input(train_batches, index)
train_pyreader.decorate_tensor_provider(data_provider)
ave_cost = 0.0
train_pyreader.start()
while True:
try:
cost = train_exe.run(fetch_list=[loss.name])
ave_cost += np.array(cost[0]).mean()
step = step + 1
if step % print_step == 0:
print("processed: [" + str(step * dev_count * 1.0 /
batch_num) + "] ave loss: [" +
str(ave_cost / print_step) + "]")
ave_cost = 0.0
if (args.save_path is not None) and (step % save_step == 0):
save_path = os.path.join(args.save_path,
"step_" + str(step))
print("Save model at step %d ... " % step)
print(time.strftime('%Y-%m-%d %H:%M:%S',
time.localtime(time.time())))
fluid.io.save_persistables(exe, save_path, train_program)
score_path = os.path.join(args.save_path,
'score.' + str(step))
test_with_pyreader(test_exe, test_program, test_pyreader,
[logits.name], score_path, val_batches,
val_batch_num, dev_count)
result_file_path = os.path.join(args.save_path,
'result.' + str(step))
evaluate(score_path, result_file_path)
except fluid.core.EOFException:
train_pyreader.reset()
break
return step, np.array(cost[0]).mean()
# train over different epoches
global_step, train_time = 0, 0.0
for epoch in six.moves.xrange(args.num_scan_data):
shuffle_train = reader.unison_shuffle(
train_data, seed=110 if ("CE_MODE_X" in os.environ) else None)
train_batches = reader.build_batches(shuffle_train, data_conf)
begin_time = time.time()
if args.use_pyreader:
global_step, last_cost = train_with_pyreader(global_step)
else:
global_step, last_cost = train_with_feed(global_step)
pass_time_cost = time.time() - begin_time
train_time += pass_time_cost
print("Pass {0}, pass_time_cost {1}"
.format(epoch, "%2.2f sec" % pass_time_cost))
# For internal continuous evaluation
if "CE_MODE_X" in os.environ:
print("kpis train_cost %f" % last_cost)
print("kpis train_duration %f" % train_time)
if __name__ == '__main__':
args = parse_args()
print_arguments(args)
train(args)
PaddleNLP/unarchived/deep_attention_matching_net/ubuntu/download_data.sh 已删除 100644 → 0
浏览文件 @ e3558f53
url=http://dam-data.cdn.bcebos.com/ubuntu.tar.gz
md5=9d7db116a040530a16f68dc0ab44e4b6
if [ ! -e ubuntu.tar.gz ]; then
wget -c $url
fi
echo "Checking md5 sum ..."
md5sum_tmp=`md5sum ubuntu.tar.gz | cut -d ' ' -f1`
if [ $md5sum_tmp != $md5 ]; then
echo "Md5sum check failed, please remove and redownload ubuntu.tar.gz"
exit 1
fi
echo "Untar ubuntu.tar.gz ..."
tar -xzvf ubuntu.tar.gz
PaddleNLP/unarchived/deep_attention_matching_net/ubuntu/test.sh 已删除 100644 → 0
浏览文件 @ e3558f53
export CUDA_VISIBLE_DEVICES=0
python -u ../test_and_evaluate.py --use_cuda \
--data_path ./data/data.pkl \
--save_path ./step_3900 \
--model_path ./models/step_3900 \
--batch_size 200 \
--vocab_size 434512 \
--emb_size 200 \
--_EOS_ 28270
PaddleNLP/unarchived/deep_attention_matching_net/ubuntu/train.sh 已删除 100644 → 0
浏览文件 @ e3558f53
export CUDA_VISIBLE_DEVICES=0
export FLAGS_eager_delete_tensor_gb=0.0
python -u ../train_and_evaluate.py --use_cuda \
--data_path ./data/data.pkl \
--word_emb_init ./data/word_embedding.pkl \
--save_path ./models \
--use_pyreader \
--batch_size 256 \
--vocab_size 434512 \
--emb_size 200 \
--_EOS_ 28270
PaddleNLP/unarchived/deep_attention_matching_net/utils/douban_evaluation.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import sys
import six
import numpy as np
from sklearn.metrics import average_precision_score
def mean_average_precision(sort_data):
#to do
count_1 = 0
sum_precision = 0
for index in six.moves.xrange(len(sort_data)):
if sort_data[index][1] == 1:
count_1 += 1
sum_precision += 1.0 * count_1 / (index + 1)
return sum_precision / count_1
def mean_reciprocal_rank(sort_data):
sort_lable = [s_d[1] for s_d in sort_data]
assert 1 in sort_lable
return 1.0 / (1 + sort_lable.index(1))
def precision_at_position_1(sort_data):
if sort_data[0][1] == 1:
return 1
else:
return 0
def recall_at_position_k_in_10(sort_data, k):
sort_lable = [s_d[1] for s_d in sort_data]
select_lable = sort_lable[:k]
return 1.0 * select_lable.count(1) / sort_lable.count(1)
def evaluation_one_session(data):
sort_data = sorted(data, key=lambda x: x[0], reverse=True)
m_a_p = mean_average_precision(sort_data)
m_r_r = mean_reciprocal_rank(sort_data)
p_1 = precision_at_position_1(sort_data)
r_1 = recall_at_position_k_in_10(sort_data, 1)
r_2 = recall_at_position_k_in_10(sort_data, 2)
r_5 = recall_at_position_k_in_10(sort_data, 5)
return m_a_p, m_r_r, p_1, r_1, r_2, r_5
def evaluate(file_path):
sum_m_a_p = 0
sum_m_r_r = 0
sum_p_1 = 0
sum_r_1 = 0
sum_r_2 = 0
sum_r_5 = 0
i = 0
total_num = 0
with open(file_path, 'r') as infile:
for line in infile:
if i % 10 == 0:
data = []
tokens = line.strip().split('\t')
data.append((float(tokens[0]), int(tokens[1])))
if i % 10 == 9:
total_num += 1
m_a_p, m_r_r, p_1, r_1, r_2, r_5 = evaluation_one_session(data)
sum_m_a_p += m_a_p
sum_m_r_r += m_r_r
sum_p_1 += p_1
sum_r_1 += r_1
sum_r_2 += r_2
sum_r_5 += r_5
i += 1
#print('total num: %s' %total_num)
#print('MAP: %s' %(1.0*sum_m_a_p/total_num))
#print('MRR: %s' %(1.0*sum_m_r_r/total_num))
#print('P@1: %s' %(1.0*sum_p_1/total_num))
return (1.0 * sum_m_a_p / total_num, 1.0 * sum_m_r_r / total_num,
1.0 * sum_p_1 / total_num, 1.0 * sum_r_1 / total_num,
1.0 * sum_r_2 / total_num, 1.0 * sum_r_5 / total_num)
if __name__ == '__main__':
result = evaluate(sys.argv[1])
for r in result:
print(r)
PaddleNLP/unarchived/deep_attention_matching_net/utils/evaluation.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import sys
import six
def get_p_at_n_in_m(data, n, m, ind):
pos_score = data[ind][0]
curr = data[ind:ind + m]
curr = sorted(curr, key=lambda x: x[0], reverse=True)
if curr[n - 1][0] <= pos_score:
return 1
return 0
def evaluate(file_path):
data = []
with open(file_path, 'r') as file:
for line in file:
line = line.strip()
tokens = line.split("\t")
if len(tokens) != 2:
continue
data.append((float(tokens[0]), int(tokens[1])))
#assert len(data) % 10 == 0
p_at_1_in_2 = 0.0
p_at_1_in_10 = 0.0
p_at_2_in_10 = 0.0
p_at_5_in_10 = 0.0
length = len(data) // 10
for i in six.moves.xrange(0, length):
ind = i * 10
assert data[ind][1] == 1
p_at_1_in_2 += get_p_at_n_in_m(data, 1, 2, ind)
p_at_1_in_10 += get_p_at_n_in_m(data, 1, 10, ind)
p_at_2_in_10 += get_p_at_n_in_m(data, 2, 10, ind)
p_at_5_in_10 += get_p_at_n_in_m(data, 5, 10, ind)
return (p_at_1_in_2 / length, p_at_1_in_10 / length, p_at_2_in_10 / length,
p_at_5_in_10 / length)
PaddleNLP/unarchived/deep_attention_matching_net/utils/layers.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import paddle.fluid as fluid
def loss(x, y, clip_value=10.0):
"""Calculate the sigmoid cross entropy with logits for input(x).
Args:
x: Variable with shape with shape [batch, dim]
y: Input label
Returns:
loss: cross entropy
logits: prediction
"""
logits = fluid.layers.fc(
input=x,
size=1,
bias_attr=fluid.ParamAttr(initializer=fluid.initializer.Constant(0.)))
loss = fluid.layers.sigmoid_cross_entropy_with_logits(x=logits, label=y)
loss = fluid.layers.reduce_mean(
fluid.layers.clip(
loss, min=-clip_value, max=clip_value))
return loss, logits
def ffn(input, d_inner_hid, d_hid, name=None):
"""Position-wise Feed-Forward Network
"""
hidden = fluid.layers.fc(input=input,
size=d_inner_hid,
num_flatten_dims=2,
param_attr=fluid.ParamAttr(name=name + '_fc.w_0'),
bias_attr=fluid.ParamAttr(
name=name + '_fc.b_0',
initializer=fluid.initializer.Constant(0.)),
act="relu")
out = fluid.layers.fc(input=hidden,
size=d_hid,
num_flatten_dims=2,
param_attr=fluid.ParamAttr(name=name + '_fc.w_1'),
bias_attr=fluid.ParamAttr(
name=name + '_fc.b_1',
initializer=fluid.initializer.Constant(0.)))
return out
def dot_product_attention(query,
key,
value,
d_key,
q_mask=None,
k_mask=None,
dropout_rate=None,
mask_cache=None):
"""Dot product layer.
Args:
query: a tensor with shape [batch, Q_time, Q_dimension]
key: a tensor with shape [batch, time, K_dimension]
value: a tensor with shape [batch, time, V_dimension]
q_lengths: a tensor with shape [batch]
k_lengths: a tensor with shape [batch]
Returns:
a tensor with shape [batch, query_time, value_dimension]
Raises:
AssertionError: if Q_dimension not equal to K_dimension when attention
type is dot.
"""
logits = fluid.layers.matmul(
x=query, y=key, transpose_y=True, alpha=d_key**(-0.5))
if (q_mask is not None) and (k_mask is not None):
if mask_cache is not None and q_mask.name in mask_cache and k_mask.name in mask_cache[
q_mask.name]:
mask, another_mask = mask_cache[q_mask.name][k_mask.name]
else:
mask = fluid.layers.matmul(x=q_mask, y=k_mask, transpose_y=True)
another_mask = fluid.layers.scale(
mask,
scale=float(2**32 - 1),
bias=float(-1),
bias_after_scale=False)
if mask_cache is not None:
if q_mask.name not in mask_cache:
mask_cache[q_mask.name] = dict()
mask_cache[q_mask.name][k_mask.name] = [mask, another_mask]
logits = mask * logits + another_mask
attention = fluid.layers.softmax(logits)
if dropout_rate:
attention = fluid.layers.dropout(
input=attention, dropout_prob=dropout_rate, is_test=False, seed=2)
atten_out = fluid.layers.matmul(x=attention, y=value)
return atten_out
def block(name,
query,
key,
value,
d_key,
q_mask=None,
k_mask=None,
is_layer_norm=True,
dropout_rate=None,
mask_cache=None):
"""
"""
att_out = dot_product_attention(
query,
key,
value,
d_key,
q_mask,
k_mask,
dropout_rate,
mask_cache=mask_cache)
y = query + att_out
if is_layer_norm:
y = fluid.layers.layer_norm(
input=y,
begin_norm_axis=len(y.shape) - 1,
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Constant(1.),
name=name + '_layer_norm.w_0'),
bias_attr=fluid.ParamAttr(
initializer=fluid.initializer.Constant(0.),
name=name + '_layer_norm.b_0'))
z = ffn(y, d_key, d_key, name)
w = y + z
if is_layer_norm:
w = fluid.layers.layer_norm(
input=w,
begin_norm_axis=len(w.shape) - 1,
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Constant(1.),
name=name + '_layer_norm.w_1'),
bias_attr=fluid.ParamAttr(
initializer=fluid.initializer.Constant(0.),
name=name + '_layer_norm.b_1'))
return w
def cnn_3d(input, out_channels_0, out_channels_1, add_relu=True):
# same padding
conv_0 = fluid.layers.conv3d(
name="conv3d_0",
input=input,
num_filters=out_channels_0,
filter_size=[3, 3, 3],
padding=[1, 1, 1],
act="elu" if add_relu else None,
param_attr=fluid.ParamAttr(initializer=fluid.initializer.Uniform(
low=-0.01, high=0.01)),
bias_attr=fluid.ParamAttr(
initializer=fluid.initializer.Constant(value=0.0)))
# same padding
pooling_0 = fluid.layers.pool3d(
input=conv_0,
pool_type="max",
pool_size=3,
pool_padding=1,
pool_stride=3)
conv_1 = fluid.layers.conv3d(
name="conv3d_1",
input=pooling_0,
num_filters=out_channels_1,
filter_size=[3, 3, 3],
padding=[1, 1, 1],
act="elu" if add_relu else None,
param_attr=fluid.ParamAttr(initializer=fluid.initializer.Uniform(
low=-0.01, high=0.01)),
bias_attr=fluid.ParamAttr(
initializer=fluid.initializer.Constant(value=0.0)))
# same padding
pooling_1 = fluid.layers.pool3d(
input=conv_1,
pool_type="max",
pool_size=3,
pool_padding=1,
pool_stride=3)
return pooling_1
PaddleNLP/unarchived/deep_attention_matching_net/utils/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import six
import numpy as np
try:
import cPickle as pickle #python 2
except ImportError as e:
import pickle #python 3
def unison_shuffle(data, seed=None):
if seed is not None:
np.random.seed(seed)
y = np.array(data[six.b('y')])
c = np.array(data[six.b('c')])
r = np.array(data[six.b('r')])
assert len(y) == len(c) == len(r)
p = np.random.permutation(len(y))
print(p)
shuffle_data = {six.b('y'): y[p], six.b('c'): c[p], six.b('r'): r[p]}
return shuffle_data
def split_c(c, split_id):
'''c is a list, example context
split_id is a integer, conf[_EOS_]
return nested list
'''
turns = [[]]
for _id in c:
if _id != split_id:
turns[-1].append(_id)
else:
turns.append([])
if turns[-1] == [] and len(turns) > 1:
turns.pop()
return turns
def normalize_length(_list, length, cut_type='tail'):
'''_list is a list or nested list, example turns/r/single turn c
cut_type is head or tail, if _list len > length is used
return a list len=length and min(read_length, length)
'''
real_length = len(_list)
if real_length == 0:
return [0] * length, 0
if real_length <= length:
if not isinstance(_list[0], list):
_list.extend([0] * (length - real_length))
else:
_list.extend([[]] * (length - real_length))
return _list, real_length
if cut_type == 'head':
return _list[:length], length
if cut_type == 'tail':
return _list[-length:], length
def produce_one_sample(data,
index,
split_id,
max_turn_num,
max_turn_len,
turn_cut_type='tail',
term_cut_type='tail'):
'''max_turn_num=10
max_turn_len=50
return y, nor_turns_nor_c, nor_r, turn_len, term_len, r_len
'''
c = data[six.b('c')][index]
r = data[six.b('r')][index][:]
y = data[six.b('y')][index]
turns = split_c(c, split_id)
#normalize turns_c length, nor_turns length is max_turn_num
nor_turns, turn_len = normalize_length(turns, max_turn_num, turn_cut_type)
nor_turns_nor_c = []
term_len = []
#nor_turn_nor_c length is max_turn_num, element is a list length is max_turn_len
for c in nor_turns:
#nor_c length is max_turn_len
nor_c, nor_c_len = normalize_length(c, max_turn_len, term_cut_type)
nor_turns_nor_c.append(nor_c)
term_len.append(nor_c_len)
nor_r, r_len = normalize_length(r, max_turn_len, term_cut_type)
return y, nor_turns_nor_c, nor_r, turn_len, term_len, r_len
def build_one_batch(data,
batch_index,
conf,
turn_cut_type='tail',
term_cut_type='tail'):
_turns = []
_tt_turns_len = []
_every_turn_len = []
_response = []
_response_len = []
_label = []
for i in six.moves.xrange(conf['batch_size']):
index = batch_index * conf['batch_size'] + i
y, nor_turns_nor_c, nor_r, turn_len, term_len, r_len = produce_one_sample(
data, index, conf['_EOS_'], conf['max_turn_num'],
conf['max_turn_len'], turn_cut_type, term_cut_type)
_label.append(y)
_turns.append(nor_turns_nor_c)
_response.append(nor_r)
_every_turn_len.append(term_len)
_tt_turns_len.append(turn_len)
_response_len.append(r_len)
return _turns, _tt_turns_len, _every_turn_len, _response, _response_len, _label
def build_one_batch_dict(data,
batch_index,
conf,
turn_cut_type='tail',
term_cut_type='tail'):
_turns, _tt_turns_len, _every_turn_len, _response, _response_len, _label = build_one_batch(
data, batch_index, conf, turn_cut_type, term_cut_type)
ans = {
'turns': _turns,
'tt_turns_len': _tt_turns_len,
'every_turn_len': _every_turn_len,
'response': _response,
'response_len': _response_len,
'label': _label
}
return ans
def build_batches(data, conf, turn_cut_type='tail', term_cut_type='tail'):
_turns_batches = []
_tt_turns_len_batches = []
_every_turn_len_batches = []
_response_batches = []
_response_len_batches = []
_label_batches = []
batch_len = len(data[six.b('y')]) // conf['batch_size']
for batch_index in six.moves.range(batch_len):
_turns, _tt_turns_len, _every_turn_len, _response, _response_len, _label = build_one_batch(
data, batch_index, conf, turn_cut_type='tail', term_cut_type='tail')
_turns_batches.append(_turns)
_tt_turns_len_batches.append(_tt_turns_len)
_every_turn_len_batches.append(_every_turn_len)
_response_batches.append(_response)
_response_len_batches.append(_response_len)
_label_batches.append(_label)
ans = {
"turns": _turns_batches,
"tt_turns_len": _tt_turns_len_batches,
"every_turn_len": _every_turn_len_batches,
"response": _response_batches,
"response_len": _response_len_batches,
"label": _label_batches
}
return ans
def make_one_batch_input(data_batches, index):
"""Split turns and return feeding data.
Args:
data_batches: All data batches
index: The index for current batch
Return:
feeding dictionary
"""
turns = np.array(data_batches["turns"][index]).astype('int64')
tt_turns_len = np.array(data_batches["tt_turns_len"][index]).astype('int64')
every_turn_len = np.array(data_batches["every_turn_len"][index]).astype(
'int64')
response = np.array(data_batches["response"][index]).astype('int64')
response_len = np.array(data_batches["response_len"][index]).astype('int64')
batch_size = turns.shape[0]
max_turn_num = turns.shape[1]
max_turn_len = turns.shape[2]
turns_list = [turns[:, i, :] for i in six.moves.xrange(max_turn_num)]
every_turn_len_list = [
every_turn_len[:, i] for i in six.moves.xrange(max_turn_num)
]
feed_list = []
for i, turn in enumerate(turns_list):
turn = np.expand_dims(turn, axis=-1)
feed_list.append(turn)
for i, turn_len in enumerate(every_turn_len_list):
turn_mask = np.ones((batch_size, max_turn_len, 1)).astype("float32")
for row in six.moves.xrange(batch_size):
turn_mask[row, turn_len[row]:, 0] = 0
feed_list.append(turn_mask)
response = np.expand_dims(response, axis=-1)
feed_list.append(response)
response_mask = np.ones((batch_size, max_turn_len, 1)).astype("float32")
for row in six.moves.xrange(batch_size):
response_mask[row, response_len[row]:, 0] = 0
feed_list.append(response_mask)
label = np.array([data_batches["label"][index]]).reshape(
[-1, 1]).astype("float32")
feed_list.append(label)
return feed_list
if __name__ == '__main__':
conf = {
"batch_size": 256,
"max_turn_num": 10,
"max_turn_len": 50,
"_EOS_": 28270,
}
with open('../ubuntu/data/data_small.pkl', 'rb') as f:
if six.PY2:
train, val, test = pickle.load(f)
else:
train, val, test = pickle.load(f, encoding="bytes")
print('load data success')
train_batches = build_batches(train, conf)
val_batches = build_batches(val, conf)
test_batches = build_batches(test, conf)
print('build batches success')
PaddleNLP/unarchived/deep_attention_matching_net/utils/util.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import six
import os
def print_arguments(args):
print('----------- Configuration Arguments -----------')
for arg, value in sorted(six.iteritems(vars(args))):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
def mkdir(path):
if not os.path.isdir(path):
mkdir(os.path.split(path)[0])
else:
return
os.mkdir(path)
def pos_encoding_init():
pass
def scaled_dot_product_attention():
pass
PaddleNLP/unarchived/language_model/gru/.run_ce.sh 已删除 100644 → 0
浏览文件 @ e3558f53
#!/bin/bash
export MKL_NUM_THREADS=1
export OMP_NUM_THREADS=1
cudaid=${language_model:=0} # use 0-th card as default
export CUDA_VISIBLE_DEVICES=$cudaid
FLAGS_benchmark=true python train.py --enable_ce | python _ce.py
cudaid=${language_model_m:=0,1,2,3} # use 0,1,2,3 card as default
export CUDA_VISIBLE_DEVICES=$cudaid
FLAGS_benchmark=true python train.py --enable_ce | python _ce.py
PaddleNLP/unarchived/language_model/gru/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
# 语言模型
以下是本例的简要目录结构及说明：
```text
.
├── README.md # 文档
├── train.py # 训练脚本
├── infer.py # 预测脚本
└── utils.py # 通用函数
```
## 简介
循环神经网络语言模型的介绍可以参阅论文[Recurrent Neural Network Regularization](https://arxiv.org/abs/1409.2329)，在本例中，我们实现了GRU-RNN语言模型。
## 训练
运行命令 `python train.py` 开始训练模型。
```python
python train.py
```
当前支持的参数可参见[train.py](./train.py) `train_net` 函数
```python
vocab, train_reader, test_reader = utils.prepare_data(
batch_size=20, # batch size
buffer_size=1000, # buffer size, default value is OK
word_freq_threshold=0) # vocabulary related parameter, and words with frequency below this value will be filtered
train(train_reader=train_reader,
vocab=vocab,
network=network,
hid_size=200, # embedding and hidden size
base_lr=1.0, # base learning rate
batch_size=20, # batch size, the same as that in prepare_data
pass_num=12, # the number of passes for training
use_cuda=True, # whether to use GPU card
parallel=False, # whether to be parallel
model_dir="model", # directory to save model
init_low_bound=-0.1, # uniform parameter initialization lower bound
init_high_bound=0.1) # uniform parameter initialization upper bound
```
## 自定义网络结构
可在[train.py](./train.py) `network` 函数中调整网络结构，当前的网络结构如下：
```python
emb = fluid.layers.embedding(input=src, size=[vocab_size, hid_size],
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Uniform(low=init_low_bound, high=init_high_bound),
learning_rate=emb_lr_x),
is_sparse=True)
fc0 = fluid.layers.fc(input=emb, size=hid_size * 3,
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Uniform(low=init_low_bound, high=init_high_bound),
learning_rate=gru_lr_x))
gru_h0 = fluid.layers.dynamic_gru(input=fc0, size=hid_size,
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Uniform(low=init_low_bound, high=init_high_bound),
learning_rate=gru_lr_x))
fc = fluid.layers.fc(input=gru_h0, size=vocab_size, act='softmax',
param_attr=fluid.ParamAttr(
initializer=fluid.initializer.Uniform(low=init_low_bound, high=init_high_bound),
learning_rate=fc_lr_x))
cost = fluid.layers.cross_entropy(input=fc, label=dst)
```
## 训练结果示例
我们在Tesla K40m单GPU卡上训练的日志如下所示
```text
epoch_1 start
step:100 ppl:771.053
step:200 ppl:449.597
step:300 ppl:642.654
step:400 ppl:458.128
step:500 ppl:510.912
step:600 ppl:451.545
step:700 ppl:364.404
step:800 ppl:324.272
step:900 ppl:360.797
step:1000 ppl:275.761
step:1100 ppl:294.599
step:1200 ppl:335.877
step:1300 ppl:185.262
step:1400 ppl:241.744
step:1500 ppl:211.507
step:1600 ppl:233.431
step:1700 ppl:298.767
step:1800 ppl:203.403
step:1900 ppl:158.828
step:2000 ppl:171.148
step:2100 ppl:280.884
epoch:1 num_steps:2104 time_cost(s):47.478780
model saved in model/epoch_1
epoch_2 start
step:100 ppl:238.099
step:200 ppl:136.527
step:300 ppl:204.184
step:400 ppl:252.886
step:500 ppl:177.377
step:600 ppl:197.688
step:700 ppl:131.650
step:800 ppl:223.906
step:900 ppl:144.785
step:1000 ppl:176.286
step:1100 ppl:148.158
step:1200 ppl:203.581
step:1300 ppl:168.208
step:1400 ppl:159.412
step:1500 ppl:114.032
step:1600 ppl:157.985
step:1700 ppl:147.743
step:1800 ppl:88.676
step:1900 ppl:141.962
step:2000 ppl:106.087
step:2100 ppl:122.709
epoch:2 num_steps:2104 time_cost(s):47.583789
model saved in model/epoch_2
...
```
## 预测
运行命令 `python infer.py model_dir start_epoch last_epoch(inclusive)` 开始预测，其中，start_epoch指定开始预测的轮次，last_epoch指定结束的轮次，例如
```python
python infer.py model 1 12 # prediction from epoch 1 to epoch 12
```
## 预测结果示例
```text
model:model/epoch_1 ppl:254.540 time_cost(s):3.29
model:model/epoch_2 ppl:177.671 time_cost(s):3.27
model:model/epoch_3 ppl:156.251 time_cost(s):3.27
model:model/epoch_4 ppl:139.036 time_cost(s):3.27
model:model/epoch_5 ppl:132.661 time_cost(s):3.27
model:model/epoch_6 ppl:130.092 time_cost(s):3.28
model:model/epoch_7 ppl:128.751 time_cost(s):3.27
model:model/epoch_8 ppl:125.411 time_cost(s):3.27
model:model/epoch_9 ppl:124.604 time_cost(s):3.28
model:model/epoch_10 ppl:124.754 time_cost(s):3.29
model:model/epoch_11 ppl:125.421 time_cost(s):3.27
model:model/epoch_12 ppl:125.676 time_cost(s):3.27
```
PaddleNLP/unarchived/language_model/gru/_ce.py 已删除 100644 → 0
浏览文件 @ e3558f53
# this file is only used for continuous evaluation test!
import os
import sys
sys.path.append(os.environ['ceroot'])
from kpi import CostKpi
from kpi import DurationKpi
imikolov_20_avg_ppl_kpi = CostKpi('imikolov_20_avg_ppl', 0.2, 0)
imikolov_20_pass_duration_kpi = DurationKpi(
'imikolov_20_pass_duration', 0.02, 0, actived=True)
imikolov_20_avg_ppl_kpi_card4 = CostKpi('imikolov_20_avg_ppl_card4', 0.2, 0)
imikolov_20_pass_duration_kpi_card4 = DurationKpi(
'imikolov_20_pass_duration_card4', 0.03, 0, actived=True)
tracking_kpis = [
imikolov_20_avg_ppl_kpi,
imikolov_20_pass_duration_kpi,
imikolov_20_avg_ppl_kpi_card4,
imikolov_20_pass_duration_kpi_card4,
]
def parse_log(log):
'''
This method should be implemented by model developers.
The suggestion:
each line in the log should be key, value, for example:
"
train_cost\t1.0
test_cost\t1.0
train_cost\t1.0
train_cost\t1.0
train_acc\t1.2
"
'''
for line in log.split('\n'):
fs = line.strip().split('\t')
print(fs)
if len(fs) == 3 and fs[0] == 'kpis':
kpi_name = fs[1]
kpi_value = float(fs[2])
yield kpi_name, kpi_value
def log_to_ce(log):
kpi_tracker = {}
for kpi in tracking_kpis:
kpi_tracker[kpi.name] = kpi
for (kpi_name, kpi_value) in parse_log(log):
print(kpi_name, kpi_value)
kpi_tracker[kpi_name].add_record(kpi_value)
kpi_tracker[kpi_name].persist()
if __name__ == '__main__':
log = sys.stdin.read()
log_to_ce(log)
PaddleNLP/unarchived/machine_reading_comprehension/data/md5sum.txt 已删除 100644 → 0
浏览文件 @ e3558f53
0ca0510fa625d35d902b73033c4ba9d8 demo.zip
dc7658b8cdf4f94b8714d130b7d15196 dureader_raw.zip
3db9a32e5a7c5375a604a70687b45479 dureader_preprocessed.zip
PaddleNLP/unarchived/machine_reading_comprehension/utils/marcov2_to_v1_tojsonl.py 已删除 100644 → 0
浏览文件 @ e3558f53
# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# 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.
import sys
import json
import pandas as pd
if __name__ == '__main__':
if len(sys.argv) != 3:
print('Usage: tojson.py <input_path> <output_path>')
exit()
infile = sys.argv[1]
outfile = sys.argv[2]
df = pd.read_json(infile)
with open(outfile, 'w') as f:
for row in df.iterrows():
f.write(row[1].to_json() + '\n')
legacy/PaddleRL/DeepQNetwork/DQN_agent.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/README_cn.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/assets/dqn.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/atari.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/atari_wrapper.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/expreplay.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/play.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/requirement.txt 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/rom_files/pong.bin 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/DeepQNetwork/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/policy_gradient/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/policy_gradient/brain.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/policy_gradient/env.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/PaddleRL/policy_gradient/run.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/README.cn.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/conv_seq2seq/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/conv_seq2seq/beamsearch.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/conv_seq2seq/download.sh 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/conv_seq2seq/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/conv_seq2seq/model.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/conv_seq2seq/preprocess.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/conv_seq2seq/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/conv_seq2seq/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/README.cn.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/avazu_data_processer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/dataset.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/images/lr_vs_dnn.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/images/wide_deep.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ctr/utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/deep_fm/README.cn.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/deep_fm/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/deep_fm/data/download.sh 已删除 100755 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/deep_fm/infer.py 已删除 100755 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/deep_fm/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/deep_fm/preprocess.py 已删除 100755 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/deep_fm/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/deep_fm/train.py 已删除 100755 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/README.cn.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/data/classification/test.txt 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/data/classification/train.txt 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/data/rank/test.txt 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/data/rank/train.txt 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/data/vocab.txt 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/images/dssm.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/images/dssm.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/images/dssm2.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/images/dssm2.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/images/dssm3.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/dssm/utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_chinese_poetry/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_chinese_poetry/README_en.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_chinese_poetry/generate.py 已删除 100755 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_chinese_poetry/preprocess.py 已删除 100755 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_chinese_poetry/reader.py 已删除 100755 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_chinese_poetry/train.py 已删除 100755 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_chinese_poetry/utils.py 已删除 100755 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_sequence_by_rnn_lm/.gitignore 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_sequence_by_rnn_lm/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_sequence_by_rnn_lm/config.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_sequence_by_rnn_lm/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_sequence_by_rnn_lm/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/generate_sequence_by_rnn_lm/utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/globally_normalized_reader/.gitignore 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/globally_normalized_reader/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/globally_normalized_reader/config.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/globally_normalized_reader/evaluate.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/globally_normalized_reader/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/globally_normalized_reader/model.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/globally_normalized_reader/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/globally_normalized_reader/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/globally_normalized_reader/vocab.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/hsigmoid/.gitignore 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/hsigmoid/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/hsigmoid/images/binary_tree.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/hsigmoid/images/network_conf.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/hsigmoid/images/path_to_1.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/hsigmoid/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/hsigmoid/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/hsigmoid/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/alexnet.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/googlenet.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/inception_v4.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/resnet.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/vgg.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/image_classification/xception.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/README_en.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/images/LambdaRank_EN.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/images/lambdarank.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/images/learning_to_rank.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/images/ranknet.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/images/ranknet_en.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/images/search_engine_example.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/lambda_rank.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/ranknet.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ltr/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/mt_with_external_memory/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/mt_with_external_memory/data_utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/mt_with_external_memory/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/mt_with_external_memory/model.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/mt_with_external_memory/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nce_cost/.gitignore 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nce_cost/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nce_cost/images/network_conf.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nce_cost/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nce_cost/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nce_cost/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nested_sequence/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nested_sequence/README_en.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/.gitignore 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/config.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/network.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/test/test_reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/test/trn_data.gz 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/neural_qa/val_and_test.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nmt_without_attention/README.cn.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nmt_without_attention/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nmt_without_attention/generate.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nmt_without_attention/images/gru.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nmt_without_attention/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/nmt_without_attention/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/config.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/decoder.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/images/503.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/images/504.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/images/505.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/images/ctc.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scene_text_recognition/utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scheduled_sampling/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scheduled_sampling/README_en.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scheduled_sampling/generate.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scheduled_sampling/images/decay.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scheduled_sampling/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scheduled_sampling/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scheduled_sampling/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/scheduled_sampling/utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/sequence_tagging_for_ner/.gitignore 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/sequence_tagging_for_ner/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/sequence_tagging_for_ner/data/test 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/sequence_tagging_for_ner/data/train 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/sequence_tagging_for_ner/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/sequence_tagging_for_ner/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/sequence_tagging_for_ner/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/sequence_tagging_for_ner/utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/README.cn.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/config/__init__.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/config/pascal_voc_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/data/label_list 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/data/prepare_voc_data.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/data_provider.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/eval.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/image_util.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/images/SSD300x300_map.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/images/ssd_network.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/images/vis_1.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/images/vis_2.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/images/vis_3.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/images/vis_4.jpg 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/vgg_ssd_net.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/ssd/visual.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/.gitignore 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/images/cnn_net.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/images/dnn_net.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/run.sh 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/text_classification/utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/README.cn.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/README.md 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/data/data.tar 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/data_processor.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/images/model_network.png 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/infer.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/infer_user.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/item_vector.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/network_conf.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/reader.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/train.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/user_vector.py 已删除 100644 → 0
浏览文件 @ e3558f53
此差异已折叠。
legacy/youtube_recall/utils.py 已删除 100644 → 0
浏览文件 @ e3558f53
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