Skip to content

M
models

PaddlePaddle / models
大约 1 年 前同步成功

通知 222
Star 6828
Fork 2962
M
models
未验证 提交 a8a58316 编写于 作者: Z zhang wenhui 提交者： GitHub
浏览文件
操作

Update paddlerec 1.8 (#4622) 

* update api 1.8

* update model 1.8
上级 57c0bf10
展开全部 隐藏空白更改
内联 并排
Showing with 128 addition and 8275 deletion
PaddleRec/ctr/Paddle_baseline_KDD2019/README.md 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Paddle_baseline_KDD2019
Paddle baseline for KDD2019 "Context-Aware Multi-Modal Transportation Recommendation"(https://dianshi.baidu.com/competition/29/question)
This repository is the demo codes for the KDD2019 "Context-Aware Multi-Modal Transportation Recommendation" competition using PaddlePaddle. It is written by python and uses PaddlePaddle to solve the task. Note that this repository is on developing and welcome everyone to contribute. The current baseline solution codes can get 0.68 - 0.69 score of online submission. As an example, my submission based on these networks programmed by PaddlePaddle is 0.6898.
The reason of the publication of this baseline codes is to encourage us to use PaddlePaddle and build the most powerful recommendation model via PaddlePaddle.
The example codes are ran on Linux, python2.7, single machine with CPU . Note that distributed train options are not provided here, if you want to learn more about this, please check more modes examples on https://github.com/PaddlePaddle/models. About the speed of training, for one epoch, 1000 batch size, it would take about 8 mins to train the whole training instances generated from raw data using SGD optimizer (it would take relatively longer using Adam optimizer).
The configuration and process of all the networks are fundamental, a lot of optimizations can be done based on them to achieve better results e.g. better cost function, more powerful feature engineering, designed model validation, NN optimization tricks...
The code is rough and from my daily use. They will be trimmed these days...
## Install PaddlePaddle
please visit the official site of PaddlePaddle(http://www.paddlepaddle.org/documentation/docs/zh/1.4/beginners_guide/install/index_cn.html)
## preprocess feature
```python
python preprocess_dense.py # change for different feature strategy
python pre_test_dense.py
```
preprocess.py and preprocess_dense.py is the code for preprocessing the raw data. Two versions are provided to deal with all sparse features and sparse plus dense features. Correspondingly, pre_process_test.py and pre_test_dense.py are the codes to preproccess test raw data. The training instances are saved in json. It is very easy to add new features. In our demo, all features are generated from provided raw data except for weather feature, which is gengerated from open weather records.
Note that the feature generated in this step need to fit in the input of the model input. Make sure we use the right version. In demo codes, The sparse plus dense features are used for network_confv6.
## build the network
main network logic is in network_confv?.py. The networks are base on fm & deep related algorithms. I try several networks and public some of them. There may be some defects in the networks but all of them are functional.
## train the network
```python
python local_train.py
```
In local_train.py and map_reader.py, I use dataset API, so we need to download the corresponding .whl package or clone codes on develop branch of PaddlePaddle. The reason to use this is the speed of feeding data is much faster.
Note that the input format feed into the network is self-defined. make sure we build the same format between training and test.
## test results
```python
python generate_test.py
python build_submit.py
```
In generate_test.py and build_submit, for convenience, I use the whole train data to train the network and test the network with provided data without label
PaddleRec/ctr/Paddle_baseline_KDD2019/args.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
import argparse
def parse_args():
parser = argparse.ArgumentParser(description="PaddlePaddle CTR example")
parser.add_argument(
'--train_data_path',
type=str,
default='./data/raw/train.txt',
help="The path of training dataset")
parser.add_argument(
'--test_data_path',
type=str,
default='./data/raw/valid.txt',
help="The path of testing dataset")
parser.add_argument(
'--batch_size',
type=int,
default=1000,
help="The size of mini-batch (default:1000)")
parser.add_argument(
'--embedding_size',
type=int,
default=16,
help="The size for embedding layer (default:10)")
parser.add_argument(
'--num_passes',
type=int,
default=10,
help="The number of passes to train (default: 10)")
parser.add_argument(
'--model_output_dir',
type=str,
default='models',
help='The path for model to store (default: models)')
parser.add_argument(
'--sparse_feature_dim',
type=int,
default=1000001,
help='sparse feature hashing space for index processing')
parser.add_argument(
'--is_local',
type=int,
default=1,
help='Local train or distributed train (default: 1)')
parser.add_argument(
'--cloud_train',
type=int,
default=0,
help='Local train or distributed train on paddlecloud (default: 0)')
parser.add_argument(
'--async_mode',
action='store_true',
default=False,
help='Whether start pserver in async mode to support ASGD')
parser.add_argument(
'--no_split_var',
action='store_true',
default=False,
help='Whether split variables into blocks when update_method is pserver')
parser.add_argument(
'--role',
type=str,
default='pserver', # trainer or pserver
help='The path for model to store (default: models)')
parser.add_argument(
'--endpoints',
type=str,
default='127.0.0.1:6000',
help='The pserver endpoints, like: 127.0.0.1:6000,127.0.0.1:6001')
parser.add_argument(
'--current_endpoint',
type=str,
default='127.0.0.1:6000',
help='The path for model to store (default: 127.0.0.1:6000)')
parser.add_argument(
'--trainer_id',
type=int,
default=0,
help='The path for model to store (default: models)')
parser.add_argument(
'--trainers',
type=int,
default=1,
help='The num of trianers, (default: 1)')
return parser.parse_args()
PaddleRec/ctr/Paddle_baseline_KDD2019/build_submit.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
import json
import csv
import io
def build():
submit_map = {}
with io.open('./submit/submit.csv', 'wb') as csv_file:
writer = csv.writer(csv_file, delimiter=',')
writer.writerow(['sid', 'recommend_mode'])
with open('./out/normed_test_session.txt', 'r') as f1:
with open('./testres/res8', 'r') as f2:
cur_session =''
for x, y in zip(f1.readlines(), f2.readlines()):
m1 = json.loads(x)
session_id = m1["session_id"]
if cur_session == '':
cur_session = session_id
transport_mode = m1["plan"]["transport_mode"]
if cur_session != session_id:
writer.writerow([str(cur_session), str(submit_map[cur_session]["transport_mode"])])
cur_session = session_id
if session_id not in submit_map:
submit_map[session_id] = {}
submit_map[session_id]["transport_mode"] = transport_mode
submit_map[session_id]["probability"] = y
#if int(submit_map[session_id]["transport_mode"]) == 0 and submit_map[session_id]["probability"] > 0.02:
#submit_map[session_id]["probability"] = 0.99
else:
if float(y) > float(submit_map[session_id]["probability"]):
submit_map[session_id]["transport_mode"] = transport_mode
submit_map[session_id]["probability"] = y
#if int(submit_map[session_id]["transport_mode"]) == 0 and submit_map[session_id]["probability"] > 0.02:
#submit_map[session_id]["transport_mode"] = 0
#submit_map[session_id]["probability"] = 0.99
writer.writerow([cur_session, submit_map[cur_session]["transport_mode"]])
if __name__ == "__main__":
build()
PaddleRec/ctr/Paddle_baseline_KDD2019/generate_test.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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 argparse
import logging
import numpy as np
# disable gpu training for this example
import os
os.environ["CUDA_VISIBLE_DEVICES"] = ""
import paddle
import paddle.fluid as fluid
logging.basicConfig(format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger("fluid")
logger.setLevel(logging.INFO)
num_context_feature = 22
def parse_args():
parser = argparse.ArgumentParser(description="PaddlePaddle DeepFM example")
parser.add_argument(
'--model_path',
type=str,
#required=True,
default='models',
help="The path of model parameters gz file")
parser.add_argument(
'--data_path',
type=str,
required=False,
help="The path of the dataset to infer")
parser.add_argument(
'--embedding_size',
type=int,
default=16,
help="The size for embedding layer (default:10)")
parser.add_argument(
'--sparse_feature_dim',
type=int,
default=1000001,
help="The size for embedding layer (default:1000001)")
parser.add_argument(
'--batch_size',
type=int,
default=1000,
help="The size of mini-batch (default:1000)")
return parser.parse_args()
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 data2tensor(data, place):
feed_dict = {}
dense = data[0]
sparse = data[1:-1]
y = data[-1]
#user_data = np.array([x[0] for x in data]).astype("float32")
#user_data = user_data.reshape([-1, 10])
#feed_dict["user_profile"] = user_data
dense_data = np.array([x[0] for x in data]).astype("float32")
dense_data = dense_data.reshape([-1, 3])
feed_dict["dense_feature"] = dense_data
for i in range(num_context_feature):
sparse_data = to_lodtensor([x[1 + i] for x in data], place)
feed_dict["context" + str(i)] = sparse_data
context_fm = to_lodtensor(
np.array([x[-2] for x in data]).astype("float32"), place)
feed_dict["context_fm"] = context_fm
y_data = np.array([x[-1] for x in data]).astype("int64")
y_data = y_data.reshape([-1, 1])
feed_dict["label"] = y_data
return feed_dict
def test():
args = parse_args()
place = fluid.CPUPlace()
test_scope = fluid.core.Scope()
# filelist = ["%s/%s" % (args.data_path, x) for x in os.listdir(args.data_path)]
from map_reader import MapDataset
map_dataset = MapDataset()
map_dataset.setup(args.sparse_feature_dim)
exe = fluid.Executor(place)
whole_filelist = ["./out/normed_test_session.txt"]
test_files = whole_filelist[int(0.0 * len(whole_filelist)):int(1.0 * len(
whole_filelist))]
epochs = 1
for i in range(epochs):
cur_model_path = os.path.join(args.model_path,
"epoch" + str(1) + ".model")
with open("./testres/res" + str(i), 'w') as r:
with fluid.scope_guard(test_scope):
[inference_program, feed_target_names, fetch_targets] = \
fluid.io.load_inference_model(cur_model_path, exe)
test_reader = map_dataset.test_reader(test_files, 1000, 100000)
k = 0
for batch_id, data in enumerate(test_reader()):
print(len(data[0]))
feed_dict = data2tensor(data, place)
loss_val, auc_val, accuracy, predict, _ = exe.run(
inference_program,
feed=feed_dict,
fetch_list=fetch_targets,
return_numpy=False)
x = np.array(predict)
for j in range(x.shape[0]):
r.write(str(x[j][1]))
r.write("\n")
if __name__ == '__main__':
test()
PaddleRec/ctr/Paddle_baseline_KDD2019/infer.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
import argparse
import logging
import numpy as np
# disable gpu training for this example
import os
os.environ["CUDA_VISIBLE_DEVICES"] = ""
import paddle
import paddle.fluid as fluid
import map_reader
from network_conf import ctr_deepfm_dataset
logging.basicConfig(format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger("fluid")
logger.setLevel(logging.INFO)
def parse_args():
parser = argparse.ArgumentParser(description="PaddlePaddle DeepFM example")
parser.add_argument(
'--model_path',
type=str,
#required=True,
default='models',
help="The path of model parameters gz file")
parser.add_argument(
'--data_path',
type=str,
required=False,
help="The path of the dataset to infer")
parser.add_argument(
'--embedding_size',
type=int,
default=16,
help="The size for embedding layer (default:10)")
parser.add_argument(
'--sparse_feature_dim',
type=int,
default=1000001,
help="The size for embedding layer (default:1000001)")
parser.add_argument(
'--batch_size',
type=int,
default=1000,
help="The size of mini-batch (default:1000)")
return parser.parse_args()
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 data2tensor(data, place):
feed_dict = {}
test_dict = {}
dense = data[0]
sparse = data[1:-1]
y = data[-1]
dense_data = np.array([x[0] for x in data]).astype("float32")
dense_data = dense_data.reshape([-1, 65])
feed_dict["user_profile"] = dense_data
for i in range(10):
sparse_data = to_lodtensor([x[1 + i] for x in data], place)
feed_dict["context" + str(i)] = sparse_data
y_data = np.array([x[-1] for x in data]).astype("int64")
y_data = y_data.reshape([-1, 1])
feed_dict["label"] = y_data
test_dict["test"] = [1]
return feed_dict, test_dict
def infer():
args = parse_args()
place = fluid.CPUPlace()
inference_scope = fluid.core.Scope()
filelist = [
"%s/%s" % (args.data_path, x) for x in os.listdir(args.data_path)
]
from map_reader import MapDataset
map_dataset = MapDataset()
map_dataset.setup(args.sparse_feature_dim)
exe = fluid.Executor(place)
whole_filelist = [
"raw_data/part-%d" % x for x in range(len(os.listdir("raw_data")))
]
#whole_filelist = ["./out/normed_train09", "./out/normed_train10", "./out/normed_train11"]
test_files = whole_filelist[int(0.0 * len(whole_filelist)):int(1.0 * len(
whole_filelist))]
# file_groups = [whole_filelist[i:i+train_thread_num] for i in range(0, len(whole_filelist), train_thread_num)]
def set_zero(var_name):
param = inference_scope.var(var_name).get_tensor()
param_array = np.zeros(param._get_dims()).astype("int64")
param.set(param_array, place)
epochs = 2
for i in range(epochs):
cur_model_path = os.path.join(args.model_path,
"epoch" + str(i + 1) + ".model")
with fluid.scope_guard(inference_scope):
[inference_program, feed_target_names, fetch_targets] = \
fluid.io.load_inference_model(cur_model_path, exe)
auc_states_names = ['_generated_var_2', '_generated_var_3']
for name in auc_states_names:
set_zero(name)
test_reader = map_dataset.infer_reader(test_files, 1000, 100000)
for batch_id, data in enumerate(test_reader()):
loss_val, auc_val, accuracy, predict, label = exe.run(
inference_program,
feed=data2tensor(data, place),
fetch_list=fetch_targets,
return_numpy=False)
#print(np.array(predict))
#x = np.array(predict)
#print(.shape)x
#print("train_pass_%d, test_pass_%d\t%f\t" % (i - 1, i, auc_val))
if __name__ == '__main__':
infer()
PaddleRec/ctr/Paddle_baseline_KDD2019/local_train.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
from __future__ import print_function
from args import parse_args
import os
import paddle.fluid as fluid
import sys
from network_confv6 import ctr_deepfm_dataset
NUM_CONTEXT_FEATURE = 22
DIM_USER_PROFILE = 10
DIM_DENSE_FEATURE = 3
PYTHON_PATH = "/home/yaoxuefeng/whls/paddle_release_home/python/bin/python" # this is mine change yours
def train():
args = parse_args()
if not os.path.isdir(args.model_output_dir):
os.mkdir(args.model_output_dir)
#set the input format for our model. Note that you need to carefully modify them when you define a new network
#user_profile = fluid.layers.data(
#name="user_profile", shape=[DIM_USER_PROFILE], dtype='int64', lod_level=1)
dense_feature = fluid.layers.data(
name="dense_feature", shape=[DIM_DENSE_FEATURE], dtype='float32')
context_feature = [
fluid.layers.data(
name="context" + str(i), shape=[1], lod_level=1, dtype="int64")
for i in range(0, NUM_CONTEXT_FEATURE)
]
context_feature_fm = fluid.layers.data(
name="context_fm", shape=[1], dtype='int64', lod_level=1)
label = fluid.layers.data(name='label', shape=[1], dtype='int64')
print("ready to network")
#self define network
loss, auc_var, batch_auc_var, accuracy, predict = ctr_deepfm_dataset(
dense_feature, context_feature, context_feature_fm, label,
args.embedding_size, args.sparse_feature_dim)
print("ready to optimize")
optimizer = fluid.optimizer.SGD(learning_rate=1e-4)
optimizer.minimize(loss)
#single machine CPU training. more options on trainig please visit PaddlePaddle site
exe = fluid.Executor(fluid.CPUPlace())
exe.run(fluid.default_startup_program())
#use dataset api for much faster speed
dataset = fluid.DatasetFactory().create_dataset()
dataset.set_use_var([dense_feature] + context_feature +
[context_feature_fm] + [label])
#self define how to process generated training insatnces in map_reader.py
pipe_command = PYTHON_PATH + " map_reader.py %d" % args.sparse_feature_dim
dataset.set_pipe_command(pipe_command)
dataset.set_batch_size(args.batch_size)
thread_num = 1
dataset.set_thread(thread_num)
#self define how to split training files for example:"split -a 2 -d -l 200000 normed_train.txt normed_train"
whole_filelist = [
"./out/normed_train%d" % x for x in range(len(os.listdir("out")))
]
whole_filelist = [
"./out/normed_train00", "./out/normed_train01", "./out/normed_train02",
"./out/normed_train03", "./out/normed_train04", "./out/normed_train05",
"./out/normed_train06", "./out/normed_train07", "./out/normed_train08",
"./out/normed_train09", "./out/normed_train10", "./out/normed_train11"
]
print("ready to epochs")
epochs = 10
for i in range(epochs):
print("start %dth epoch" % i)
dataset.set_filelist(whole_filelist[:int(len(whole_filelist))])
#print the informations you want by setting fetch_list and fetch_info
exe.train_from_dataset(
program=fluid.default_main_program(),
dataset=dataset,
fetch_list=[auc_var, accuracy, predict, label],
fetch_info=["auc", "accuracy", "predict", "label"],
debug=False)
model_dir = os.path.join(args.model_output_dir,
'/epoch' + str(i + 1) + ".model")
sys.stderr.write("epoch%d finished" % (i + 1))
#save model
fluid.io.save_inference_model(
model_dir,
[dense_feature.name] + [x.name for x in context_feature] +
[context_feature_fm.name] + [label.name],
[loss, auc_var, accuracy, predict, label], exe)
if __name__ == '__main__':
train()
PaddleRec/ctr/Paddle_baseline_KDD2019/map_reader.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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 paddle.fluid.incubate.data_generator as dg
class MapDataset(dg.MultiSlotDataGenerator):
def setup(self, sparse_feature_dim):
self.profile_length = 65
self.dense_length = 3
#feature names
self.dense_feature_list = ["distance", "price", "eta"]
self.pid_list = ["pid"]
self.query_feature_list = ["weekday", "hour", "o1", "o2", "d1", "d2"]
self.plan_feature_list = ["transport_mode"]
self.rank_feature_list = ["plan_rank", "whole_rank", "price_rank", "eta_rank", "distance_rank"]
self.rank_whole_pic_list = ["mode_rank1", "mode_rank2", "mode_rank3", "mode_rank4",
"mode_rank5"]
self.weather_feature_list = ["max_temp", "min_temp", "wea", "wind"]
self.hash_dim = 1000001
self.train_idx_ = 2000000
#carefully set if you change the features
self.categorical_range_ = range(0, 22)
#process one instance
def _process_line(self, line):
instance = json.loads(line)
"""
profile = instance["profile"]
len_profile = len(profile)
if len_profile >= 10:
user_profile_feature = profile[0:10]
else:
profile.extend([0]*(10-len_profile))
user_profile_feature = profile
if len(profile) > 1 or (len(profile) == 1 and profile[0] != 0):
for p in profile:
if p >= 1 and p <= 65:
user_profile_feature[p - 1] = 1
"""
context_feature = []
context_feature_fm = []
dense_feature = [0] * self.dense_length
plan = instance["plan"]
for i, val in enumerate(self.dense_feature_list):
dense_feature[i] = plan[val]
if (instance["pid"] == ""):
instance["pid"] = 0
query = instance["query"]
weather_dic = instance["weather"]
for fea in self.pid_list:
context_feature.append([hash(fea + str(instance[fea])) % self.hash_dim])
context_feature_fm.append(hash(fea + str(instance[fea])) % self.hash_dim)
for fea in self.query_feature_list:
context_feature.append([hash(fea + str(query[fea])) % self.hash_dim])
context_feature_fm.append(hash(fea + str(query[fea])) % self.hash_dim)
for fea in self.plan_feature_list:
context_feature.append([hash(fea + str(plan[fea])) % self.hash_dim])
context_feature_fm.append(hash(fea + str(plan[fea])) % self.hash_dim)
for fea in self.rank_feature_list:
context_feature.append([hash(fea + str(instance[fea])) % self.hash_dim])
context_feature_fm.append(hash(fea + str(instance[fea])) % self.hash_dim)
for fea in self.rank_whole_pic_list:
context_feature.append([hash(fea + str(instance[fea])) % self.hash_dim])
context_feature_fm.append(hash(fea + str(instance[fea])) % self.hash_dim)
for fea in self.weather_feature_list:
context_feature.append([hash(fea + str(weather_dic[fea])) % self.hash_dim])
context_feature_fm.append(hash(fea + str(weather_dic[fea])) % self.hash_dim)
label = [int(instance["label"])]
return dense_feature, context_feature, context_feature_fm, label
def infer_reader(self, filelist, batch, buf_size):
print(filelist)
def local_iter():
for fname in filelist:
with open(fname.strip(), "r") as fin:
for line in fin:
dense_feature, sparse_feature, sparse_feature_fm, label = self._process_line(line)
yield [dense_feature] + sparse_feature + [sparse_feature_fm] + [label]
import paddle
batch_iter = paddle.batch(
paddle.reader.shuffle(
local_iter, buf_size=buf_size),
batch_size=batch)
return batch_iter
#generat inputs for testing
def test_reader(self, filelist, batch, buf_size):
print(filelist)
def local_iter():
for fname in filelist:
with open(fname.strip(), "r") as fin:
for line in fin:
dense_feature, sparse_feature, sparse_feature_fm, label = self._process_line(line)
yield [dense_feature] + sparse_feature + [sparse_feature_fm] + [label]
import paddle
batch_iter = paddle.batch(
paddle.reader.buffered(
local_iter, size=buf_size),
batch_size=batch)
return batch_iter
#generate inputs for trainig
def generate_sample(self, line):
def data_iter():
dense_feature, sparse_feature, sparse_feature_fm, label = self._process_line(line)
#feature_name = ["user_profile"]
feature_name = []
feature_name.append("dense_feature")
for idx in self.categorical_range_:
feature_name.append("context" + str(idx))
feature_name.append("context_fm")
feature_name.append("label")
yield zip(feature_name, [dense_feature] + sparse_feature + [sparse_feature_fm] + [label])
return data_iter
if __name__ == "__main__":
map_dataset = MapDataset()
map_dataset.setup(int(sys.argv[1]))
map_dataset.run_from_stdin()
PaddleRec/ctr/Paddle_baseline_KDD2019/network_confv6.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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
import math
user_profile_dim = 65
dense_feature_dim = 3
def ctr_deepfm_dataset(dense_feature, context_feature, context_feature_fm, label,
embedding_size, sparse_feature_dim):
def dense_fm_layer(input, emb_dict_size, factor_size, fm_param_attr):
first_order = fluid.layers.fc(input=input, size=1)
emb_table = fluid.layers.create_parameter(shape=[emb_dict_size, factor_size],
dtype='float32', attr=fm_param_attr)
input_mul_factor = fluid.layers.matmul(input, emb_table)
input_mul_factor_square = fluid.layers.square(input_mul_factor)
input_square = fluid.layers.square(input)
factor_square = fluid.layers.square(emb_table)
input_square_mul_factor_square = fluid.layers.matmul(input_square, factor_square)
second_order = 0.5 * (input_mul_factor_square - input_square_mul_factor_square)
return first_order, second_order
dense_fm_param_attr = fluid.param_attr.ParamAttr(name="DenseFeatFactors",
initializer=fluid.initializer.Normal(
scale=1 / math.sqrt(dense_feature_dim)))
dense_fm_first, dense_fm_second = dense_fm_layer(
dense_feature, dense_feature_dim, 16, dense_fm_param_attr)
def sparse_fm_layer(input, emb_dict_size, factor_size, fm_param_attr):
first_embeddings = fluid.layers.embedding(
input=input, dtype='float32', size=[emb_dict_size, 1], is_sparse=True)
first_order = fluid.layers.sequence_pool(input=first_embeddings, pool_type='sum')
nonzero_embeddings = fluid.layers.embedding(
input=input, dtype='float32', size=[emb_dict_size, factor_size],
param_attr=fm_param_attr, is_sparse=True)
summed_features_emb = fluid.layers.sequence_pool(input=nonzero_embeddings, pool_type='sum')
summed_features_emb_square = fluid.layers.square(summed_features_emb)
squared_features_emb = fluid.layers.square(nonzero_embeddings)
squared_sum_features_emb = fluid.layers.sequence_pool(
input=squared_features_emb, pool_type='sum')
second_order = 0.5 * (summed_features_emb_square - squared_sum_features_emb)
return first_order, second_order
sparse_fm_param_attr = fluid.param_attr.ParamAttr(name="SparseFeatFactors",
initializer=fluid.initializer.Normal(
scale=1 / math.sqrt(sparse_feature_dim)))
#data = fluid.layers.data(name='ids', shape=[1], dtype='float32')
sparse_fm_first, sparse_fm_second = sparse_fm_layer(
context_feature_fm, sparse_feature_dim, 16, sparse_fm_param_attr)
def embedding_layer(input):
return fluid.layers.embedding(
input=input,
is_sparse=True,
# you need to patch https://github.com/PaddlePaddle/Paddle/pull/14190
# if you want to set is_distributed to True
is_distributed=False,
size=[sparse_feature_dim, embedding_size],
param_attr=fluid.ParamAttr(name="SparseFeatFactors",
initializer=fluid.initializer.Uniform()))
sparse_embed_seq = list(map(embedding_layer, context_feature))
concated_ori = fluid.layers.concat(sparse_embed_seq + [dense_feature], axis=1)
concated = fluid.layers.batch_norm(input=concated_ori, name="bn", epsilon=1e-4)
deep = deep_net(concated)
predict = fluid.layers.fc(input=[deep, sparse_fm_first, sparse_fm_second, dense_fm_first, dense_fm_second], size=2, act="softmax",
param_attr=fluid.ParamAttr(initializer=fluid.initializer.Normal(
scale=1 / math.sqrt(deep.shape[1])), learning_rate=0.01))
#similarity_norm = fluid.layers.sigmoid(fluid.layers.clip(predict, min=-15.0, max=15.0), name="similarity_norm")
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.reduce_sum(cost)
accuracy = fluid.layers.accuracy(input=predict, label=label)
auc_var, batch_auc_var, auc_states = \
fluid.layers.auc(input=predict, label=label, num_thresholds=2 ** 12, slide_steps=20)
return avg_cost, auc_var, batch_auc_var, accuracy, predict
def deep_net(concated, lr_x=0.0001):
fc_layers_input = [concated]
fc_layers_size = [400, 400, 400]
fc_layers_act = ["relu"] * (len(fc_layers_size))
for i in range(len(fc_layers_size)):
fc = fluid.layers.fc(
input=fc_layers_input[-1],
size=fc_layers_size[i],
act=fc_layers_act[i],
param_attr=fluid.ParamAttr(learning_rate=lr_x * 0.5))
fc_layers_input.append(fc)
#w_res = fluid.layers.create_parameter(shape=[353, 16], dtype='float32', name="w_res")
#high_path = fluid.layers.matmul(concated, w_res)
#return fluid.layers.elementwise_add(high_path, fc_layers_input[-1])
return fc_layers_input[-1]
PaddleRec/ctr/Paddle_baseline_KDD2019/networks/network_conf.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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
import math
user_profile_dim = 65
num_context = 25
dim_fm_vector = 16
dim_concated = user_profile_dim + dim_fm_vector * (num_context)
def ctr_deepfm_dataset(user_profile, context_feature, label,
embedding_size, sparse_feature_dim):
def embedding_layer(input):
return fluid.layers.embedding(
input=input,
is_sparse=True,
# you need to patch https://github.com/PaddlePaddle/Paddle/pull/14190
# if you want to set is_distributed to True
is_distributed=False,
size=[sparse_feature_dim, embedding_size],
param_attr=fluid.ParamAttr(name="SparseFeatFactors",
initializer=fluid.initializer.Uniform()))
sparse_embed_seq = list(map(embedding_layer, context_feature))
w = fluid.layers.create_parameter(
shape=[65, 65], dtype='float32',
name="w_fm")
user_profile_emb = fluid.layers.matmul(user_profile, w)
concated_ori = fluid.layers.concat(sparse_embed_seq + [user_profile_emb], axis=1)
concated = fluid.layers.batch_norm(input=concated_ori, name="bn", epsilon=1e-4)
deep = deep_net(concated)
linear_term, second_term = fm(concated, dim_concated, 8) #depend on the number of context feature
predict = fluid.layers.fc(input=[deep, linear_term, second_term], size=2, act="softmax",
param_attr=fluid.ParamAttr(initializer=fluid.initializer.Normal(
scale=1 / math.sqrt(deep.shape[1])), learning_rate=0.01))
#similarity_norm = fluid.layers.sigmoid(fluid.layers.clip(predict, min=-15.0, max=15.0), name="similarity_norm")
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.reduce_sum(cost)
accuracy = fluid.layers.accuracy(input=predict, label=label)
auc_var, batch_auc_var, auc_states = \
fluid.layers.auc(input=predict, label=label, num_thresholds=2 ** 12, slide_steps=20)
return avg_cost, auc_var, batch_auc_var, accuracy, predict
def deep_net(concated, lr_x=0.0001):
fc_layers_input = [concated]
fc_layers_size = [128, 64, 32, 16]
fc_layers_act = ["relu"] * (len(fc_layers_size))
for i in range(len(fc_layers_size)):
fc = fluid.layers.fc(
input=fc_layers_input[-1],
size=fc_layers_size[i],
act=fc_layers_act[i],
param_attr=fluid.ParamAttr(learning_rate=lr_x * 0.5))
fc_layers_input.append(fc)
return fc_layers_input[-1]
def fm(concated, emb_dict_size, factor_size, lr_x=0.0001):
linear_term = fluid.layers.fc(input=concated, size=8, act=None, param_attr=fluid.ParamAttr(learning_rate=lr_x))
emb_table = fluid.layers.create_parameter(shape=[emb_dict_size, factor_size],
dtype='float32')
input_mul_factor = fluid.layers.matmul(concated, emb_table)
input_mul_factor_square = fluid.layers.square(input_mul_factor)
input_square = fluid.layers.square(concated)
factor_square = fluid.layers.square(emb_table)
input_square_mul_factor_square = fluid.layers.matmul(input_square, factor_square)
second_term = 0.5 * (input_mul_factor_square - input_square_mul_factor_square)
return linear_term, second_term
PaddleRec/ctr/Paddle_baseline_KDD2019/networks/network_confv4.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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
import math
user_profile_dim = 65
slot_1 = [0, 1, 2, 3, 4, 5]
slot_2 = [6]
slot_3 = [7, 8, 9, 10, 11]
slot_4 = [12, 13, 14, 15, 16]
slot_5 = [17, 18, 19, 20]
num_context = 25
num_slots_pair = 5
dim_fm_vector = 16
dim_concated = user_profile_dim + dim_fm_vector * (num_context + num_slots_pair)
def ctr_deepfm_dataset(user_profile, dense_feature, context_feature, label,
embedding_size, sparse_feature_dim):
def embedding_layer(input):
return fluid.layers.embedding(
input=input,
is_sparse=True,
# you need to patch https://github.com/PaddlePaddle/Paddle/pull/14190
# if you want to set is_distributed to True
is_distributed=False,
size=[sparse_feature_dim, embedding_size],
param_attr=fluid.ParamAttr(name="SparseFeatFactors",
initializer=fluid.initializer.Uniform()))
sparse_embed_seq = list(map(embedding_layer, context_feature))
w = fluid.layers.create_parameter(
shape=[65, 65], dtype='float32',
name="w_fm")
user_emb_list = []
user_profile_emb = fluid.layers.matmul(user_profile, w)
user_emb_list.append(user_profile_emb)
user_emb_list.append(dense_feature)
w1 = fluid.layers.create_parameter(shape=[65, dim_fm_vector], dtype='float32', name="w_1")
w2 = fluid.layers.create_parameter(shape=[65, dim_fm_vector], dtype='float32', name="w_2")
w3 = fluid.layers.create_parameter(shape=[65, dim_fm_vector], dtype='float32', name="w_3")
w4 = fluid.layers.create_parameter(shape=[65, dim_fm_vector], dtype='float32', name="w_4")
w5 = fluid.layers.create_parameter(shape=[65, dim_fm_vector], dtype='float32', name="w_5")
user_profile_emb_1 = fluid.layers.matmul(user_profile, w1)
user_profile_emb_2 = fluid.layers.matmul(user_profile, w2)
user_profile_emb_3 = fluid.layers.matmul(user_profile, w3)
user_profile_emb_4 = fluid.layers.matmul(user_profile, w4)
user_profile_emb_5 = fluid.layers.matmul(user_profile, w5)
sparse_embed_seq_1 = embedding_layer(context_feature[slot_1[0]])
sparse_embed_seq_2 = embedding_layer(context_feature[slot_2[0]])
sparse_embed_seq_3 = embedding_layer(context_feature[slot_3[0]])
sparse_embed_seq_4 = embedding_layer(context_feature[slot_4[0]])
sparse_embed_seq_5 = embedding_layer(context_feature[slot_5[0]])
for i in slot_1[1:-1]:
sparse_embed_seq_1 = fluid.layers.elementwise_add(sparse_embed_seq_1, embedding_layer(context_feature[i]))
for i in slot_2[1:-1]:
sparse_embed_seq_2 = fluid.layers.elementwise_add(sparse_embed_seq_2, embedding_layer(context_feature[i]))
for i in slot_3[1:-1]:
sparse_embed_seq_3 = fluid.layers.elementwise_add(sparse_embed_seq_3, embedding_layer(context_feature[i]))
for i in slot_4[1:-1]:
sparse_embed_seq_4 = fluid.layers.elementwise_add(sparse_embed_seq_4, embedding_layer(context_feature[i]))
for i in slot_5[1:-1]:
sparse_embed_seq_5 = fluid.layers.elementwise_add(sparse_embed_seq_5, embedding_layer(context_feature[i]))
ele_product_1 = fluid.layers.elementwise_mul(user_profile_emb_1, sparse_embed_seq_1)
user_emb_list.append(ele_product_1)
ele_product_2 = fluid.layers.elementwise_mul(user_profile_emb_2, sparse_embed_seq_2)
user_emb_list.append(ele_product_2)
ele_product_3 = fluid.layers.elementwise_mul(user_profile_emb_3, sparse_embed_seq_3)
user_emb_list.append(ele_product_3)
ele_product_4 = fluid.layers.elementwise_mul(user_profile_emb_4, sparse_embed_seq_4)
user_emb_list.append(ele_product_4)
ele_product_5 = fluid.layers.elementwise_mul(user_profile_emb_5, sparse_embed_seq_5)
user_emb_list.append(ele_product_5)
ffm_1 = fluid.layers.reduce_sum(ele_product_1, dim=1, keep_dim=True)
ffm_2 = fluid.layers.reduce_sum(ele_product_2, dim=1, keep_dim=True)
ffm_3 = fluid.layers.reduce_sum(ele_product_3, dim=1, keep_dim=True)
ffm_4 = fluid.layers.reduce_sum(ele_product_4, dim=1, keep_dim=True)
ffm_5 = fluid.layers.reduce_sum(ele_product_5, dim=1, keep_dim=True)
concated_ori = fluid.layers.concat(sparse_embed_seq + user_emb_list, axis=1)
concated = fluid.layers.batch_norm(input=concated_ori, name="bn", epsilon=1e-4)
deep = deep_net(concated)
linear_term, second_term = fm(concated, dim_concated, 8) #depend on the number of context feature
predict = fluid.layers.fc(input=[deep, linear_term, second_term, ffm_1, ffm_2, ffm_3, ffm_4, ffm_5], size=2, act="softmax",
param_attr=fluid.ParamAttr(initializer=fluid.initializer.Normal(
scale=1 / math.sqrt(deep.shape[1])), learning_rate=0.01))
#similarity_norm = fluid.layers.sigmoid(fluid.layers.clip(predict, min=-15.0, max=15.0), name="similarity_norm")
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.reduce_sum(cost)
accuracy = fluid.layers.accuracy(input=predict, label=label)
auc_var, batch_auc_var, auc_states = \
fluid.layers.auc(input=predict, label=label, num_thresholds=2 ** 12, slide_steps=20)
return avg_cost, auc_var, batch_auc_var, accuracy, predict
def deep_net(concated, lr_x=0.0001):
fc_layers_input = [concated]
fc_layers_size = [256, 128, 64, 32, 16]
fc_layers_act = ["relu"] * (len(fc_layers_size))
for i in range(len(fc_layers_size)):
fc = fluid.layers.fc(
input=fc_layers_input[-1],
size=fc_layers_size[i],
act=fc_layers_act[i],
param_attr=fluid.ParamAttr(learning_rate=lr_x * 0.5))
fc_layers_input.append(fc)
w_res = fluid.layers.create_parameter(shape=[dim_concated, 16], dtype='float32', name="w_res")
high_path = fluid.layers.matmul(concated, w_res)
return fluid.layers.elementwise_add(high_path, fc_layers_input[-1])
#return fc_layers_input[-1]
def fm(concated, emb_dict_size, factor_size, lr_x=0.0001):
linear_term = fluid.layers.fc(input=concated, size=8, act=None, param_attr=fluid.ParamAttr(learning_rate=lr_x))
emb_table = fluid.layers.create_parameter(shape=[emb_dict_size, factor_size],
dtype='float32')
input_mul_factor = fluid.layers.matmul(concated, emb_table)
input_mul_factor_square = fluid.layers.square(input_mul_factor)
input_square = fluid.layers.square(concated)
factor_square = fluid.layers.square(emb_table)
input_square_mul_factor_square = fluid.layers.matmul(input_square, factor_square)
second_term = 0.5 * (input_mul_factor_square - input_square_mul_factor_square)
return linear_term, second_term
\ No newline at end of file
PaddleRec/ctr/Paddle_baseline_KDD2019/networks/network_confv6.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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
import math
user_profile_dim = 65
dense_feature_dim = 3
def ctr_deepfm_dataset(dense_feature, context_feature, context_feature_fm, label,
embedding_size, sparse_feature_dim):
def dense_fm_layer(input, emb_dict_size, factor_size, fm_param_attr):
first_order = fluid.layers.fc(input=input, size=1)
emb_table = fluid.layers.create_parameter(shape=[emb_dict_size, factor_size],
dtype='float32', attr=fm_param_attr)
input_mul_factor = fluid.layers.matmul(input, emb_table)
input_mul_factor_square = fluid.layers.square(input_mul_factor)
input_square = fluid.layers.square(input)
factor_square = fluid.layers.square(emb_table)
input_square_mul_factor_square = fluid.layers.matmul(input_square, factor_square)
second_order = 0.5 * (input_mul_factor_square - input_square_mul_factor_square)
return first_order, second_order
dense_fm_param_attr = fluid.param_attr.ParamAttr(name="DenseFeatFactors",
initializer=fluid.initializer.Normal(
scale=1 / math.sqrt(dense_feature_dim)))
dense_fm_first, dense_fm_second = dense_fm_layer(
dense_feature, dense_feature_dim, 16, dense_fm_param_attr)
def sparse_fm_layer(input, emb_dict_size, factor_size, fm_param_attr):
first_embeddings = fluid.layers.embedding(
input=input, dtype='float32', size=[emb_dict_size, 1], is_sparse=True)
first_order = fluid.layers.sequence_pool(input=first_embeddings, pool_type='sum')
nonzero_embeddings = fluid.layers.embedding(
input=input, dtype='float32', size=[emb_dict_size, factor_size],
param_attr=fm_param_attr, is_sparse=True)
summed_features_emb = fluid.layers.sequence_pool(input=nonzero_embeddings, pool_type='sum')
summed_features_emb_square = fluid.layers.square(summed_features_emb)
squared_features_emb = fluid.layers.square(nonzero_embeddings)
squared_sum_features_emb = fluid.layers.sequence_pool(
input=squared_features_emb, pool_type='sum')
second_order = 0.5 * (summed_features_emb_square - squared_sum_features_emb)
return first_order, second_order
sparse_fm_param_attr = fluid.param_attr.ParamAttr(name="SparseFeatFactors",
initializer=fluid.initializer.Normal(
scale=1 / math.sqrt(sparse_feature_dim)))
#data = fluid.layers.data(name='ids', shape=[1], dtype='float32')
sparse_fm_first, sparse_fm_second = sparse_fm_layer(
context_feature_fm, sparse_feature_dim, 16, sparse_fm_param_attr)
def embedding_layer(input):
return fluid.layers.embedding(
input=input,
is_sparse=True,
# you need to patch https://github.com/PaddlePaddle/Paddle/pull/14190
# if you want to set is_distributed to True
is_distributed=False,
size=[sparse_feature_dim, embedding_size],
param_attr=fluid.ParamAttr(name="SparseFeatFactors",
initializer=fluid.initializer.Uniform()))
sparse_embed_seq = list(map(embedding_layer, context_feature))
concated_ori = fluid.layers.concat(sparse_embed_seq + [dense_feature], axis=1)
concated = fluid.layers.batch_norm(input=concated_ori, name="bn", epsilon=1e-4)
deep = deep_net(concated)
predict = fluid.layers.fc(input=[deep, sparse_fm_first, sparse_fm_second, dense_fm_first, dense_fm_second], size=2, act="softmax",
param_attr=fluid.ParamAttr(initializer=fluid.initializer.Normal(
scale=1 / math.sqrt(deep.shape[1])), learning_rate=0.01))
#similarity_norm = fluid.layers.sigmoid(fluid.layers.clip(predict, min=-15.0, max=15.0), name="similarity_norm")
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.reduce_sum(cost)
accuracy = fluid.layers.accuracy(input=predict, label=label)
auc_var, batch_auc_var, auc_states = \
fluid.layers.auc(input=predict, label=label, num_thresholds=2 ** 12, slide_steps=20)
return avg_cost, auc_var, batch_auc_var, accuracy, predict
def deep_net(concated, lr_x=0.0001):
fc_layers_input = [concated]
fc_layers_size = [400, 400, 400]
fc_layers_act = ["relu"] * (len(fc_layers_size))
for i in range(len(fc_layers_size)):
fc = fluid.layers.fc(
input=fc_layers_input[-1],
size=fc_layers_size[i],
act=fc_layers_act[i],
param_attr=fluid.ParamAttr(learning_rate=lr_x * 0.5))
fc_layers_input.append(fc)
#w_res = fluid.layers.create_parameter(shape=[353, 16], dtype='float32', name="w_res")
#high_path = fluid.layers.matmul(concated, w_res)
#return fluid.layers.elementwise_add(high_path, fc_layers_input[-1])
return fc_layers_input[-1]
\ No newline at end of file
PaddleRec/ctr/Paddle_baseline_KDD2019/pre_process_test.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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, sys, time, random, csv, datetime, json
import pandas as pd
import numpy as np
import argparse
import logging
import time
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger("preprocess")
logger.setLevel(logging.INFO)
TEST_QUERIES_PATH = "./data_set_phase1/test_queries.csv"
TEST_PLANS_PATH = "./data_set_phase1/test_plans.csv"
TRAIN_CLICK_PATH = "./data_set_phase1/train_clicks.csv"
PROFILES_PATH = "./data_set_phase1/profiles.csv"
OUT_NORM_TEST_PATH = "./out/normed_test_session.txt"
OUT_RAW_TEST_PATH = "./out/test_session.txt"
O1_MIN = 115.47
O1_MAX = 117.29
O2_MIN = 39.46
O2_MAX = 40.97
D1_MIN = 115.44
D1_MAX = 117.37
D2_MIN = 39.46
D2_MAX = 40.96
SCALE_OD = 0.02
DISTANCE_MIN = 1.0
DISTANCE_MAX = 225864.0
THRESHOLD_DIS = 40000.0
SCALE_DIS = 500
PRICE_MIN = 200.0
PRICE_MAX = 92300.0
THRESHOLD_PRICE = 20000
SCALE_PRICE = 100
ETA_MIN = 1.0
ETA_MAX = 72992.0
THRESHOLD_ETA = 10800.0
SCALE_ETA = 120
def build_norm_feature():
with open(OUT_NORM_TEST_PATH, 'w') as nf:
with open(OUT_RAW_TEST_PATH, 'r') as f:
for line in f:
cur_map = json.loads(line)
if cur_map["plan"]["distance"] > THRESHOLD_DIS:
cur_map["plan"]["distance"] = int(THRESHOLD_DIS)
elif cur_map["plan"]["distance"] > 0:
cur_map["plan"]["distance"] = int(cur_map["plan"]["distance"] / SCALE_DIS)
if cur_map["plan"]["price"] and cur_map["plan"]["price"] > THRESHOLD_PRICE:
cur_map["plan"]["price"] = int(THRESHOLD_PRICE)
elif not cur_map["plan"]["price"] or cur_map["plan"]["price"] < 0:
cur_map["plan"]["price"] = 0
else:
cur_map["plan"]["price"] = int(cur_map["plan"]["price"] / SCALE_PRICE)
if cur_map["plan"]["eta"] > THRESHOLD_ETA:
cur_map["plan"]["eta"] = int(THRESHOLD_ETA)
elif cur_map["plan"]["eta"] > 0:
cur_map["plan"]["eta"] = int(cur_map["plan"]["eta"] / SCALE_ETA)
# o1
if cur_map["query"]["o1"] > O1_MAX:
cur_map["query"]["o1"] = int((O1_MAX - O1_MIN) / SCALE_OD + 1)
elif cur_map["query"]["o1"] < O1_MIN:
cur_map["query"]["o1"] = 0
else:
cur_map["query"]["o1"] = int((cur_map["query"]["o1"] - O1_MIN) / 0.02)
# o2
if cur_map["query"]["o2"] > O2_MAX:
cur_map["query"]["o2"] = int((O2_MAX - O2_MIN) / SCALE_OD + 1)
elif cur_map["query"]["o2"] < O2_MIN:
cur_map["query"]["o2"] = 0
else:
cur_map["query"]["o2"] = int((cur_map["query"]["o2"] - O2_MIN) / 0.02)
# d1
if cur_map["query"]["d1"] > D1_MAX:
cur_map["query"]["d1"] = int((D1_MAX - D1_MIN) / SCALE_OD + 1)
elif cur_map["query"]["d1"] < D1_MIN:
cur_map["query"]["d1"] = 0
else:
cur_map["query"]["d1"] = int((cur_map["query"]["d1"] - D1_MIN) / SCALE_OD)
# d2
if cur_map["query"]["d2"] > D2_MAX:
cur_map["query"]["d2"] = int((D2_MAX - D2_MIN) / SCALE_OD + 1)
elif cur_map["query"]["d2"] < D2_MIN:
cur_map["query"]["d2"] = 0
else:
cur_map["query"]["d2"] = int((cur_map["query"]["d2"] - D2_MIN) / SCALE_OD)
cur_json_instance = json.dumps(cur_map)
nf.write(cur_json_instance + '\n')
def preprocess():
"""
Construct the train data indexed by session id and mode id jointly. Convert some of the raw features (user profile,
od pair, req time, click time, eta, price, distance, transport mode) to one-hot ids used for
embedding. We split the one-hot features into two categories: user feature and context feature for
better understanding of FM algorithm.
Note that the user profile is already provided by one-hot encoded form, we convert it back to the
ids for unity with the context feature and easily using of PaddlePaddle embedding layer. Given the
train clicks data, we label each train instance with 1 or 0 depend on if this instance is clicked or
not.
:return:
"""
train_data_dict = {}
with open("./weather.json", 'r') as f:
weather_dict = json.load(f)
with open(TEST_QUERIES_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
train_index_list = []
for k, line in enumerate(csv_reader):
if k == 0: continue
if line[0] == "": continue
if line[1] == "":
train_index_list.append(line[0] + "_0")
else:
train_index_list.append(line[0] + "_" + line[1])
train_index = line[0]
train_data_dict[train_index] = {}
train_data_dict[train_index]["pid"] = line[1]
train_data_dict[train_index]["query"] = {}
reqweekday = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%w")
reqhour = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%H")
date_key = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%m-%d")
train_data_dict[train_index]["weather"] = {}
train_data_dict[train_index]["weather"].update({"max_temp": weather_dict[date_key]["max_temp"]})
train_data_dict[train_index]["weather"].update({"min_temp": weather_dict[date_key]["min_temp"]})
train_data_dict[train_index]["weather"].update({"wea": weather_dict[date_key]["weather"]})
train_data_dict[train_index]["weather"].update({"wind": weather_dict[date_key]["wind"]})
train_data_dict[train_index]["query"].update({"weekday":reqweekday})
train_data_dict[train_index]["query"].update({"hour":reqhour})
o = line[3].split(',')
o_first = o[0]
o_second = o[1]
train_data_dict[train_index]["query"].update({"o1":float(o_first)})
train_data_dict[train_index]["query"].update({"o2":float(o_second)})
d = line[4].split(',')
d_first = d[0]
d_second = d[1]
train_data_dict[train_index]["query"].update({"d1":float(d_first)})
train_data_dict[train_index]["query"].update({"d2":float(d_second)})
plan_map = {}
plan_data = pd.read_csv(TEST_PLANS_PATH)
for index, row in plan_data.iterrows():
plans_str = row['plans']
plans_list = json.loads(plans_str)
session_id = str(row['sid'])
# train_data_dict[session_id]["plans"] = []
plan_map[session_id] = plans_list
profile_map = {}
with open(PROFILES_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
for k, line in enumerate(csv_reader):
if k == 0: continue
profile_map[line[0]] = [i for i in range(len(line)) if line[i] == "1.0"]
session_click_map = {}
with open(TRAIN_CLICK_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
for k, line in enumerate(csv_reader):
if k == 0: continue
if line[0] == "" or line[1] == "" or line[2] == "":
continue
session_click_map[line[0]] = line[2]
#return train_data_dict, profile_map, session_click_map, plan_map
generate_sparse_features(train_data_dict, profile_map, session_click_map, plan_map)
def generate_sparse_features(train_data_dict, profile_map, session_click_map, plan_map):
if not os.path.isdir("./out/"):
os.mkdir("./out/")
with open(os.path.join("./out/", "test_session.txt"), 'w') as f_train:
for session_id, plan_list in plan_map.items():
if session_id not in train_data_dict:
continue
cur_map = train_data_dict[session_id]
cur_map["session_id"] = session_id
if cur_map["pid"] != "":
cur_map["profile"] = profile_map[cur_map["pid"]]
else:
cur_map["profile"] = [0]
del cur_map["pid"]
whole_rank = 0
for plan in plan_list:
whole_rank += 1
cur_map["mode_rank" + str(whole_rank)] = plan["transport_mode"]
if whole_rank < 5:
for r in range(whole_rank + 1, 6):
cur_map["mode_rank" + str(r)] = -1
cur_map["whole_rank"] = whole_rank
flag_click = False
rank = 1
price_list = []
eta_list = []
distance_list = []
for plan in plan_list:
if not plan["price"]:
price_list.append(0)
else:
price_list.append(int(plan["price"]))
eta_list.append(int(plan["eta"]))
distance_list.append(int(plan["distance"]))
price_list.sort(reverse=False)
eta_list.sort(reverse=False)
distance_list.sort(reverse=False)
for plan in plan_list:
if plan["price"] and int(plan["price"]) == price_list[0]:
cur_map["mode_min_price"] = plan["transport_mode"]
if plan["price"] and int(plan["price"]) == price_list[-1]:
cur_map["mode_max_price"] = plan["transport_mode"]
if int(plan["eta"]) == eta_list[0]:
cur_map["mode_min_eta"] = plan["transport_mode"]
if int(plan["eta"]) == eta_list[-1]:
cur_map["mode_max_eta"] = plan["transport_mode"]
if int(plan["distance"]) == distance_list[0]:
cur_map["mode_min_distance"] = plan["transport_mode"]
if int(plan["distance"]) == distance_list[-1]:
cur_map["mode_max_distance"] = plan["transport_mode"]
if "mode_min_price" not in cur_map:
cur_map["mode_min_price"] = -1
if "mode_max_price" not in cur_map:
cur_map["mode_max_price"] = -1
for plan in plan_list:
cur_price = int(plan["price"]) if plan["price"] else 0
cur_eta = int(plan["eta"])
cur_distance = int(plan["distance"])
cur_map["price_rank"] = price_list.index(cur_price) + 1
cur_map["eta_rank"] = eta_list.index(cur_eta) + 1
cur_map["distance_rank"] = distance_list.index(cur_distance) + 1
if ("transport_mode" in plan) and (session_id in session_click_map) and (
int(plan["transport_mode"]) == int(session_click_map[session_id])):
cur_map["plan"] = plan
cur_map["label"] = 1
flag_click = True
# print("label is 1")
else:
cur_map["plan"] = plan
cur_map["label"] = 0
cur_map["plan_rank"] = rank
rank += 1
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
cur_map["plan"]["distance"] = -1
cur_map["plan"]["price"] = -1
cur_map["plan"]["eta"] = -1
cur_map["plan"]["transport_mode"] = 0
cur_map["plan_rank"] = 0
cur_map["price_rank"] = 0
cur_map["eta_rank"] = 0
cur_map["plan_rank"] = 0
cur_map["label"] = 1
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
build_norm_feature()
if __name__ == "__main__":
preprocess()
\ No newline at end of file
PaddleRec/ctr/Paddle_baseline_KDD2019/pre_test_dense.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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, sys, time, random, csv, datetime, json
import pandas as pd
import numpy as np
import argparse
import logging
import time
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger("preprocess")
logger.setLevel(logging.INFO)
TRAIN_QUERIES_PATH = "./data_set_phase1/test_queries.csv"
TRAIN_PLANS_PATH = "./data_set_phase1/test_plans.csv"
TRAIN_CLICK_PATH = "./data_set_phase1/train_clicks.csv"
PROFILES_PATH = "./data_set_phase1/profiles.csv"
O1_MIN = 115.47
O1_MAX = 117.29
O2_MIN = 39.46
O2_MAX = 40.97
D1_MIN = 115.44
D1_MAX = 117.37
D2_MIN = 39.46
D2_MAX = 40.96
DISTANCE_MIN = 1.0
DISTANCE_MAX = 225864.0
THRESHOLD_DIS = 200000.0
PRICE_MIN = 200.0
PRICE_MAX = 92300.0
THRESHOLD_PRICE = 20000
ETA_MIN = 1.0
ETA_MAX = 72992.0
THRESHOLD_ETA = 10800.0
def build_norm_feature():
with open("./out/normed_test_session.txt", 'w') as nf:
with open("./out/test_session.txt", 'r') as f:
for line in f:
cur_map = json.loads(line)
cur_map["plan"]["distance"] = (cur_map["plan"]["distance"] - DISTANCE_MIN) / (DISTANCE_MAX - DISTANCE_MIN)
if cur_map["plan"]["price"]:
cur_map["plan"]["price"] = (cur_map["plan"]["price"] - PRICE_MIN) / (PRICE_MAX - PRICE_MIN)
else:
cur_map["plan"]["price"] = 0.0
cur_map["plan"]["eta"] = (cur_map["plan"]["eta"] - ETA_MIN) / (ETA_MAX - ETA_MIN)
cur_json_instance = json.dumps(cur_map)
nf.write(cur_json_instance + '\n')
def preprocess():
"""
Construct the train data indexed by session id and mode id jointly. Convert all the raw features (user profile,
od pair, req time, click time, eta, price, distance, transport mode) to one-hot ids used for
embedding. We split the one-hot features into two categories: user feature and context feature for
better understanding of FFM algorithm.
Note that the user profile is already provided by one-hot encoded form, we convert it back to the
ids for unity with the context feature and easily using of PaddlePaddle embedding layer. Given the
train clicks data, we label each train instance with 1 or 0 depend on if this instance is clicked or
not.
:return:
"""
#args = parse_args()
train_data_dict = {}
with open("./weather.json", 'r') as f:
weather_dict = json.load(f)
with open(TRAIN_QUERIES_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
train_index_list = []
for k, line in enumerate(csv_reader):
if k == 0: continue
if line[0] == "": continue
if line[1] == "":
train_index_list.append(line[0] + "_0")
else:
train_index_list.append(line[0] + "_" + line[1])
train_index = line[0]
train_data_dict[train_index] = {}
train_data_dict[train_index]["pid"] = line[1]
train_data_dict[train_index]["query"] = {}
reqweekday = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%w")
reqhour = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%H")
date_key = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%m-%d")
train_data_dict[train_index]["weather"] = {}
train_data_dict[train_index]["weather"].update({"max_temp": weather_dict[date_key]["max_temp"]})
train_data_dict[train_index]["weather"].update({"min_temp": weather_dict[date_key]["min_temp"]})
train_data_dict[train_index]["weather"].update({"wea": weather_dict[date_key]["weather"]})
train_data_dict[train_index]["weather"].update({"wind": weather_dict[date_key]["wind"]})
train_data_dict[train_index]["query"].update({"weekday":reqweekday})
train_data_dict[train_index]["query"].update({"hour":reqhour})
o = line[3].split(',')
o_first = o[0]
o_second = o[1]
train_data_dict[train_index]["query"].update({"o1":float(o_first)})
train_data_dict[train_index]["query"].update({"o2":float(o_second)})
d = line[4].split(',')
d_first = d[0]
d_second = d[1]
train_data_dict[train_index]["query"].update({"d1":float(d_first)})
train_data_dict[train_index]["query"].update({"d2":float(d_second)})
plan_map = {}
plan_data = pd.read_csv(TRAIN_PLANS_PATH)
for index, row in plan_data.iterrows():
plans_str = row['plans']
plans_list = json.loads(plans_str)
session_id = str(row['sid'])
# train_data_dict[session_id]["plans"] = []
plan_map[session_id] = plans_list
profile_map = {}
with open(PROFILES_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
for k, line in enumerate(csv_reader):
if k == 0: continue
profile_map[line[0]] = [i for i in range(len(line)) if line[i] == "1.0"]
session_click_map = {}
with open(TRAIN_CLICK_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
for k, line in enumerate(csv_reader):
if k == 0: continue
if line[0] == "" or line[1] == "" or line[2] == "":
continue
session_click_map[line[0]] = line[2]
#return train_data_dict, profile_map, session_click_map, plan_map
generate_sparse_features(train_data_dict, profile_map, session_click_map, plan_map)
def generate_sparse_features(train_data_dict, profile_map, session_click_map, plan_map):
if not os.path.isdir("./out/"):
os.mkdir("./out/")
with open(os.path.join("./out/", "test_session.txt"), 'w') as f_train:
for session_id, plan_list in plan_map.items():
if session_id not in train_data_dict:
continue
cur_map = train_data_dict[session_id]
cur_map["session_id"] = session_id
if cur_map["pid"] != "":
cur_map["profile"] = profile_map[cur_map["pid"]]
else:
cur_map["profile"] = [0]
# del cur_map["pid"]
whole_rank = 0
for plan in plan_list:
whole_rank += 1
cur_map["mode_rank" + str(whole_rank)] = plan["transport_mode"]
if whole_rank < 5:
for r in range(whole_rank + 1, 6):
cur_map["mode_rank" + str(r)] = -1
cur_map["whole_rank"] = whole_rank
rank = 1
price_list = []
eta_list = []
distance_list = []
for plan in plan_list:
if not plan["price"]:
price_list.append(0)
else:
price_list.append(int(plan["price"]))
eta_list.append(int(plan["eta"]))
distance_list.append(int(plan["distance"]))
price_list.sort(reverse=False)
eta_list.sort(reverse=False)
distance_list.sort(reverse=False)
for plan in plan_list:
if plan["price"] and int(plan["price"]) == price_list[0]:
cur_map["mode_min_price"] = plan["transport_mode"]
if plan["price"] and int(plan["price"]) == price_list[-1]:
cur_map["mode_max_price"] = plan["transport_mode"]
if int(plan["eta"]) == eta_list[0]:
cur_map["mode_min_eta"] = plan["transport_mode"]
if int(plan["eta"]) == eta_list[-1]:
cur_map["mode_max_eta"] = plan["transport_mode"]
if int(plan["distance"]) == distance_list[0]:
cur_map["mode_min_distance"] = plan["transport_mode"]
if int(plan["distance"]) == distance_list[-1]:
cur_map["mode_max_distance"] = plan["transport_mode"]
if "mode_min_price" not in cur_map:
cur_map["mode_min_price"] = -1
if "mode_max_price" not in cur_map:
cur_map["mode_max_price"] = -1
for plan in plan_list:
cur_price = int(plan["price"]) if plan["price"] else 0
cur_eta = int(plan["eta"])
cur_distance = int(plan["distance"])
cur_map["price_rank"] = price_list.index(cur_price) + 1
cur_map["eta_rank"] = eta_list.index(cur_eta) + 1
cur_map["distance_rank"] = distance_list.index(cur_distance) + 1
if ("transport_mode" in plan) and (session_id in session_click_map) and (
int(plan["transport_mode"]) == int(session_click_map[session_id])):
cur_map["plan"] = plan
cur_map["label"] = 1
else:
cur_map["plan"] = plan
cur_map["label"] = 0
cur_map["plan_rank"] = rank
rank += 1
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
cur_map["plan"]["distance"] = -1
cur_map["plan"]["price"] = -1
cur_map["plan"]["eta"] = -1
cur_map["plan"]["transport_mode"] = 0
cur_map["plan_rank"] = 0
cur_map["price_rank"] = 0
cur_map["eta_rank"] = 0
cur_map["plan_rank"] = 0
cur_map["label"] = 1
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
build_norm_feature()
if __name__ == "__main__":
preprocess()
\ No newline at end of file
PaddleRec/ctr/Paddle_baseline_KDD2019/preprocess.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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, sys, time, random, csv, datetime, json
import pandas as pd
import numpy as np
import argparse
import logging
import time
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger("preprocess")
logger.setLevel(logging.INFO)
TRAIN_QUERIES_PATH = "./data_set_phase1/train_queries.csv"
TRAIN_PLANS_PATH = "./data_set_phase1/train_plans.csv"
TRAIN_CLICK_PATH = "./data_set_phase1/train_clicks.csv"
PROFILES_PATH = "./data_set_phase1/profiles.csv"
OUT_NORM_TRAIN_PATH = "./out/normed_train.txt"
OUT_RAW_TRAIN_PATH = "./out/train.txt"
OUT_DIR = "./out"
O1_MIN = 115.47
O1_MAX = 117.29
O2_MIN = 39.46
O2_MAX = 40.97
D1_MIN = 115.44
D1_MAX = 117.37
D2_MIN = 39.46
D2_MAX = 40.96
SCALE_OD = 0.02
DISTANCE_MIN = 1.0
DISTANCE_MAX = 225864.0
THRESHOLD_DIS = 40000.0
SCALE_DIS = 500
PRICE_MIN = 200.0
PRICE_MAX = 92300.0
THRESHOLD_PRICE = 20000
SCALE_PRICE = 100
ETA_MIN = 1.0
ETA_MAX = 72992.0
THRESHOLD_ETA = 10800.0
SCALE_ETA = 120
def build_norm_feature():
with open(OUT_NORM_TRAIN_PATH, 'w') as nf:
with open(OUT_RAW_TRAIN_PATH, 'r') as f:
for line in f:
cur_map = json.loads(line)
if cur_map["plan"]["distance"] > THRESHOLD_DIS:
cur_map["plan"]["distance"] = int(THRESHOLD_DIS)
elif cur_map["plan"]["distance"] > 0:
cur_map["plan"]["distance"] = int(cur_map["plan"]["distance"] / SCALE_DIS)
if cur_map["plan"]["price"] and cur_map["plan"]["price"] > THRESHOLD_PRICE:
cur_map["plan"]["price"] = int(THRESHOLD_PRICE)
elif not cur_map["plan"]["price"] or cur_map["plan"]["price"] < 0:
cur_map["plan"]["price"] = 0
else:
cur_map["plan"]["price"] = int(cur_map["plan"]["price"] / SCALE_PRICE)
if cur_map["plan"]["eta"] > THRESHOLD_ETA:
cur_map["plan"]["eta"] = int(THRESHOLD_ETA)
elif cur_map["plan"]["eta"] > 0:
cur_map["plan"]["eta"] = int(cur_map["plan"]["eta"] / SCALE_ETA)
# o1
if cur_map["query"]["o1"] > O1_MAX:
cur_map["query"]["o1"] = int((O1_MAX - O1_MIN) / SCALE_OD + 1)
elif cur_map["query"]["o1"] < O1_MIN:
cur_map["query"]["o1"] = 0
else:
cur_map["query"]["o1"] = int((cur_map["query"]["o1"] - O1_MIN) / 0.02)
# o2
if cur_map["query"]["o2"] > O2_MAX:
cur_map["query"]["o2"] = int((O2_MAX - O2_MIN) / SCALE_OD + 1)
elif cur_map["query"]["o2"] < O2_MIN:
cur_map["query"]["o2"] = 0
else:
cur_map["query"]["o2"] = int((cur_map["query"]["o2"] - O2_MIN) / 0.02)
# d1
if cur_map["query"]["d1"] > D1_MAX:
cur_map["query"]["d1"] = int((D1_MAX - D1_MIN) / SCALE_OD + 1)
elif cur_map["query"]["d1"] < D1_MIN:
cur_map["query"]["d1"] = 0
else:
cur_map["query"]["d1"] = int((cur_map["query"]["d1"] - D1_MIN) / SCALE_OD)
# d2
if cur_map["query"]["d2"] > D2_MAX:
cur_map["query"]["d2"] = int((D2_MAX - D2_MIN) / SCALE_OD + 1)
elif cur_map["query"]["d2"] < D2_MIN:
cur_map["query"]["d2"] = 0
else:
cur_map["query"]["d2"] = int((cur_map["query"]["d2"] - D2_MIN) / SCALE_OD)
cur_json_instance = json.dumps(cur_map)
nf.write(cur_json_instance + '\n')
def preprocess():
"""
Construct the train data indexed by session id and mode id jointly. Convert all the raw features (user profile,
od pair, req time, click time, eta, price, distance, transport mode) to one-hot ids used for
embedding. We split the one-hot features into two categories: user feature and context feature for
better understanding of FM algorithm.
Note that the user profile is already provided by one-hot encoded form, we treat it as embedded vector
for unity with the context feature and easily using of PaddlePaddle embedding layer. Given the
train clicks data, we label each train instance with 1 or 0 depend on if this instance is clicked or
not include non-click case.
:return:
"""
train_data_dict = {}
with open(TRAIN_QUERIES_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
train_index_list = []
for k, line in enumerate(csv_reader):
if k == 0: continue
if line[0] == "": continue
if line[1] == "":
train_index_list.append(line[0] + "_0")
else:
train_index_list.append(line[0] + "_" + line[1])
train_index = line[0]
train_data_dict[train_index] = {}
train_data_dict[train_index]["pid"] = line[1]
train_data_dict[train_index]["query"] = {}
reqweekday = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%w")
reqhour = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%H")
train_data_dict[train_index]["query"].update({"weekday":reqweekday})
train_data_dict[train_index]["query"].update({"hour":reqhour})
o = line[3].split(',')
o_first = o[0]
o_second = o[1]
train_data_dict[train_index]["query"].update({"o1":float(o_first)})
train_data_dict[train_index]["query"].update({"o2":float(o_second)})
d = line[4].split(',')
d_first = d[0]
d_second = d[1]
train_data_dict[train_index]["query"].update({"d1":float(d_first)})
train_data_dict[train_index]["query"].update({"d2":float(d_second)})
plan_map = {}
plan_data = pd.read_csv(TRAIN_PLANS_PATH)
for index, row in plan_data.iterrows():
plans_str = row['plans']
plans_list = json.loads(plans_str)
session_id = str(row['sid'])
# train_data_dict[session_id]["plans"] = []
plan_map[session_id] = plans_list
profile_map = {}
with open(PROFILES_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
for k, line in enumerate(csv_reader):
if k == 0: continue
profile_map[line[0]] = [i for i in range(len(line)) if line[i] == "1.0"]
session_click_map = {}
with open(TRAIN_CLICK_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
for k, line in enumerate(csv_reader):
if k == 0: continue
if line[0] == "" or line[1] == "" or line[2] == "":
continue
session_click_map[line[0]] = line[2]
#return train_data_dict, profile_map, session_click_map, plan_map
generate_sparse_features(train_data_dict, profile_map, session_click_map, plan_map)
def generate_sparse_features(train_data_dict, profile_map, session_click_map, plan_map):
if not os.path.isdir(OUT_DIR):
os.mkdir(OUT_DIR)
with open(os.path.join("./out/", "train.txt"), 'w') as f_train:
for session_id, plan_list in plan_map.items():
if session_id not in train_data_dict:
continue
cur_map = train_data_dict[session_id]
if cur_map["pid"] != "":
cur_map["profile"] = profile_map[cur_map["pid"]]
else:
cur_map["profile"] = [0]
del cur_map["pid"]
whole_rank = 0
for plan in plan_list:
whole_rank += 1
cur_map["whole_rank"] = whole_rank
flag_click = False
rank = 1
for plan in plan_list:
if ("transport_mode" in plan) and (session_id in session_click_map) and (
int(plan["transport_mode"]) == int(session_click_map[session_id])):
cur_map["plan"] = plan
cur_map["label"] = 1
flag_click = True
# print("label is 1")
else:
cur_map["plan"] = plan
cur_map["label"] = 0
cur_map["rank"] = rank
rank += 1
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
if not flag_click:
cur_map["plan"]["distance"] = -1
cur_map["plan"]["price"] = -1
cur_map["plan"]["eta"] = -1
cur_map["plan"]["transport_mode"] = 0
cur_map["rank"] = 0
cur_map["label"] = 1
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
else:
cur_map["plan"]["distance"] = -1
cur_map["plan"]["price"] = -1
cur_map["plan"]["eta"] = -1
cur_map["plan"]["transport_mode"] = 0
cur_map["rank"] = 0
cur_map["label"] = 0
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
build_norm_feature()
if __name__ == "__main__":
preprocess()
PaddleRec/ctr/Paddle_baseline_KDD2019/preprocess_dense.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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, random, csv, datetime, json
import pandas as pd
import numpy as np
import argparse
import logging
import time
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger("preprocess")
logger.setLevel(logging.INFO)
TRAIN_QUERIES_PATH = "./data_set_phase1/train_queries.csv"
TRAIN_PLANS_PATH = "./data_set_phase1/train_plans.csv"
TRAIN_CLICK_PATH = "./data_set_phase1/train_clicks.csv"
PROFILES_PATH = "./data_set_phase1/profiles.csv"
OUT_DIR = "./out"
ORI_TRAIN_PATH = "train.txt"
NORM_TRAIN_PATH = "normed_train.txt"
#variable to control the ratio of positive and negative instance of transmode 0 which is original label of no click
THRESHOLD_LABEL = 0.5
O1_MIN = 115.47
O1_MAX = 117.29
O2_MIN = 39.46
O2_MAX = 40.97
D1_MIN = 115.44
D1_MAX = 117.37
D2_MIN = 39.46
D2_MAX = 40.96
DISTANCE_MIN = 1.0
DISTANCE_MAX = 225864.0
THRESHOLD_DIS = 200000.0
PRICE_MIN = 200.0
PRICE_MAX = 92300.0
THRESHOLD_PRICE = 20000
ETA_MIN = 1.0
ETA_MAX = 72992.0
THRESHOLD_ETA = 10800.0
def build_norm_feature():
with open(os.path.join(OUT_DIR, NORM_TRAIN_PATH), 'w') as nf:
with open(os.path.join(OUT_DIR, ORI_TRAIN_PATH), 'r') as f:
for line in f:
cur_map = json.loads(line)
cur_map["plan"]["distance"] = (cur_map["plan"]["distance"] - DISTANCE_MIN) / (DISTANCE_MAX - DISTANCE_MIN)
if cur_map["plan"]["price"]:
cur_map["plan"]["price"] = (cur_map["plan"]["price"] - PRICE_MIN) / (PRICE_MAX - PRICE_MIN)
else:
cur_map["plan"]["price"] = 0.0
cur_map["plan"]["eta"] = (cur_map["plan"]["eta"] - ETA_MIN) / (ETA_MAX - ETA_MIN)
cur_json_instance = json.dumps(cur_map)
nf.write(cur_json_instance + '\n')
def preprocess():
"""
Construct the train data indexed by session id and mode id jointly. Convert all the raw features (user profile,
od pair, req time, click time, eta, price, distance, transport mode) to one-hot ids used for
embedding. We split the one-hot features into two categories: user feature and context feature for
better understanding of FM algorithm.
Note that the user profile is already provided by one-hot encoded form, we treat it as embedded vector
for unity with the context feature and easily using of PaddlePaddle embedding layer. Given the
train clicks data, we label each train instance with 1 or 0 depend on if this instance is clicked or
not include non-click case. To Be Changed
:return:
"""
train_data_dict = {}
with open("./weather.json", 'r') as f:
weather_dict = json.load(f)
with open(TRAIN_QUERIES_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
train_index_list = []
for k, line in enumerate(csv_reader):
if k == 0: continue
if line[0] == "": continue
if line[1] == "":
train_index_list.append(line[0] + "_0")
else:
train_index_list.append(line[0] + "_" + line[1])
train_index = line[0]
train_data_dict[train_index] = {}
train_data_dict[train_index]["pid"] = line[1]
train_data_dict[train_index]["query"] = {}
train_data_dict[train_index]["weather"] = {}
reqweekday = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%w")
reqhour = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%H")
# weather related features, no big use, maybe more detailed weather information is better
date_key = datetime.datetime.strptime(line[2], '%Y-%m-%d %H:%M:%S').strftime("%m-%d")
train_data_dict[train_index]["weather"] = {}
train_data_dict[train_index]["weather"].update({"max_temp": weather_dict[date_key]["max_temp"]})
train_data_dict[train_index]["weather"].update({"min_temp": weather_dict[date_key]["min_temp"]})
train_data_dict[train_index]["weather"].update({"wea": weather_dict[date_key]["weather"]})
train_data_dict[train_index]["weather"].update({"wind": weather_dict[date_key]["wind"]})
train_data_dict[train_index]["query"].update({"weekday":reqweekday})
train_data_dict[train_index]["query"].update({"hour":reqhour})
o = line[3].split(',')
o_first = o[0]
o_second = o[1]
train_data_dict[train_index]["query"].update({"o1":float(o_first)})
train_data_dict[train_index]["query"].update({"o2":float(o_second)})
d = line[4].split(',')
d_first = d[0]
d_second = d[1]
train_data_dict[train_index]["query"].update({"d1":float(d_first)})
train_data_dict[train_index]["query"].update({"d2":float(d_second)})
plan_map = {}
plan_data = pd.read_csv(TRAIN_PLANS_PATH)
for index, row in plan_data.iterrows():
plans_str = row['plans']
plans_list = json.loads(plans_str)
session_id = str(row['sid'])
# train_data_dict[session_id]["plans"] = []
plan_map[session_id] = plans_list
profile_map = {}
with open(PROFILES_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
for k, line in enumerate(csv_reader):
if k == 0: continue
profile_map[line[0]] = [i for i in range(len(line)) if line[i] == "1.0"]
session_click_map = {}
with open(TRAIN_CLICK_PATH, 'r') as f:
csv_reader = csv.reader(f, delimiter=',')
for k, line in enumerate(csv_reader):
if k == 0: continue
if line[0] == "" or line[1] == "" or line[2] == "":
continue
session_click_map[line[0]] = line[2]
#return train_data_dict, profile_map, session_click_map, plan_map
generate_sparse_features(train_data_dict, profile_map, session_click_map, plan_map)
def generate_sparse_features(train_data_dict, profile_map, session_click_map, plan_map):
if not os.path.isdir(OUT_DIR):
os.mkdir(OUT_DIR)
with open(os.path.join(OUT_DIR, ORI_TRAIN_PATH), 'w') as f_train:
for session_id, plan_list in plan_map.items():
if session_id not in train_data_dict:
continue
cur_map = train_data_dict[session_id]
if cur_map["pid"] != "":
cur_map["profile"] = profile_map[cur_map["pid"]]
else:
cur_map["profile"] = [0]
#rank information related feature
whole_rank = 0
for plan in plan_list:
whole_rank += 1
cur_map["mode_rank" + str(whole_rank)] = plan["transport_mode"]
if whole_rank < 5:
for r in range(whole_rank + 1, 6):
cur_map["mode_rank" + str(r)] = -1
cur_map["whole_rank"] = whole_rank
flag_click = False
rank = 1
price_list = []
eta_list = []
distance_list = []
for plan in plan_list:
if not plan["price"]:
price_list.append(0)
else:
price_list.append(int(plan["price"]))
eta_list.append(int(plan["eta"]))
distance_list.append(int(plan["distance"]))
price_list.sort(reverse=False)
eta_list.sort(reverse=False)
distance_list.sort(reverse=False)
for plan in plan_list:
if plan["price"] and int(plan["price"]) == price_list[0]:
cur_map["mode_min_price"] = plan["transport_mode"]
if plan["price"] and int(plan["price"]) == price_list[-1]:
cur_map["mode_max_price"] = plan["transport_mode"]
if int(plan["eta"]) == eta_list[0]:
cur_map["mode_min_eta"] = plan["transport_mode"]
if int(plan["eta"]) == eta_list[-1]:
cur_map["mode_max_eta"] = plan["transport_mode"]
if int(plan["distance"]) == distance_list[0]:
cur_map["mode_min_distance"] = plan["transport_mode"]
if int(plan["distance"]) == distance_list[-1]:
cur_map["mode_max_distance"] = plan["transport_mode"]
if "mode_min_price" not in cur_map:
cur_map["mode_min_price"] = -1
if "mode_max_price" not in cur_map:
cur_map["mode_max_price"] = -1
for plan in plan_list:
if ("transport_mode" in plan) and (session_id in session_click_map) and (
int(plan["transport_mode"]) == int(session_click_map[session_id])):
flag_click = True
if flag_click:
for plan in plan_list:
cur_price = int(plan["price"]) if plan["price"] else 0
cur_eta = int(plan["eta"])
cur_distance = int(plan["distance"])
cur_map["price_rank"] = price_list.index(cur_price) + 1
cur_map["eta_rank"] = eta_list.index(cur_eta) + 1
cur_map["distance_rank"] = distance_list.index(cur_distance) + 1
if ("transport_mode" in plan) and (session_id in session_click_map) and (
int(plan["transport_mode"]) == int(session_click_map[session_id])):
cur_map["plan"] = plan
cur_map["label"] = 1
else:
cur_map["plan"] = plan
cur_map["label"] = 0
cur_map["plan_rank"] = rank
rank += 1
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
cur_map["plan"] = {}
#since we define a new ctr task from original task, we use a basic way to generate instances of transport mode 0.
#There should be a optimal strategy to generate instances of transport mode 0
if not flag_click:
cur_map["plan"]["distance"] = -1
cur_map["plan"]["price"] = -1
cur_map["plan"]["eta"] = -1
cur_map["plan"]["transport_mode"] = 0
cur_map["plan_rank"] = 0
cur_map["price_rank"] = 0
cur_map["eta_rank"] = 0
cur_map["distance_rank"] = 0
cur_map["label"] = 1
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
else:
if random.random() < THRESHOLD_LABEL:
cur_map["plan"]["distance"] = -1
cur_map["plan"]["price"] = -1
cur_map["plan"]["eta"] = -1
cur_map["plan"]["transport_mode"] = 0
cur_map["plan_rank"] = 0
cur_map["price_rank"] = 0
cur_map["eta_rank"] = 0
cur_map["distance_rank"] = 0
cur_map["label"] = 0
cur_json_instance = json.dumps(cur_map)
f_train.write(cur_json_instance + '\n')
build_norm_feature()
if __name__ == "__main__":
preprocess()
PaddleRec/ctr/Paddle_baseline_KDD2019/weather.json 已删除 100644 → 0
浏览文件 @ 57c0bf10
{"10-01": {"max_temp": "24", "min_temp": "12", "weather": "q", "wind": "45"}, "10-02": {"max_temp": "24", "min_temp": "11", "weather": "q", "wind": "12"}, "10-03": {"max_temp": "25", "min_temp": "10", "weather": "q", "wind": "12"}, "10-04": {"max_temp": "25", "min_temp": "12", "weather": "q", "wind": "12"}, "10-05": {"max_temp": "24", "min_temp": "14", "weather": "dy", "wind": "12"}, "10-06": {"max_temp": "20", "min_temp": "8", "weather": "q", "wind": "45"}, "10-07": {"max_temp": "21", "min_temp": "7", "weather": "q", "wind": "12"}, "10-08": {"max_temp": "21", "min_temp": "8", "weather": "dy", "wind": "12"}, "10-09": {"max_temp": "15", "min_temp": "4", "weather": "dyq", "wind": "45"}, "10-10": {"max_temp": "17", "min_temp": "4", "weather": "dyq", "wind": "12"}, "10-11": {"max_temp": "18", "min_temp": "5", "weather": "qdy", "wind": "12"}, "10-12": {"max_temp": "20", "min_temp": "5", "weather": "dyq", "wind": "12"}, "10-13": {"max_temp": "20", "min_temp": "8", "weather": "dy", "wind": "12"}, "10-14": {"max_temp": "21", "min_temp": "10", "weather": "dy", "wind": "12"}, "10-15": {"max_temp": "17", "min_temp": "11", "weather": "xq", "wind": "12"}, "10-16": {"max_temp": "17", "min_temp": "7", "weather": "dyq", "wind": "12"}, "10-17": {"max_temp": "17", "min_temp": "5", "weather": "q", "wind": "12"}, "10-18": {"max_temp": "18", "min_temp": "5", "weather": "q", "wind": "12"}, "10-19": {"max_temp": "19", "min_temp": "7", "weather": "dy", "wind": "12"}, "10-20": {"max_temp": "18", "min_temp": "7", "weather": "dy", "wind": "12"}, "10-21": {"max_temp": "18", "min_temp": "7", "weather": "dy", "wind": "12"}, "10-22": {"max_temp": "19", "min_temp": "5", "weather": "dyq", "wind": "12"}, "10-23": {"max_temp": "19", "min_temp": "4", "weather": "q", "wind": "34"}, "10-24": {"max_temp": "20", "min_temp": "6", "weather": "qdy", "wind": "12"}, "10-25": {"max_temp": "15", "min_temp": "8", "weather": "dy", "wind": "12"}, "10-26": {"max_temp": "14", "min_temp": "3", "weather": "q", "wind": "45"}, "10-27": {"max_temp": "17", "min_temp": "5", "weather": "dy", "wind": "12"}, "10-28": {"max_temp": "17", "min_temp": "4", "weather": "dyq", "wind": "45"}, "10-29": {"max_temp": "15", "min_temp": "3", "weather": "q", "wind": "34"}, "10-30": {"max_temp": "16", "min_temp": "1", "weather": "q", "wind": "12"}, "10-31": {"max_temp": "17", "min_temp": "3", "weather": "q", "wind": "12"}, "11-01": {"max_temp": "17", "min_temp": "3", "weather": "q", "wind": "12"}, "11-02": {"max_temp": "18", "min_temp": "4", "weather": "q", "wind": "12"}, "11-03": {"max_temp": "16", "min_temp": "6", "weather": "dy", "wind": "12"}, "11-04": {"max_temp": "10", "min_temp": "2", "weather": "xydy", "wind": "34"}, "11-05": {"max_temp": "10", "min_temp": "2", "weather": "dy", "wind": "12"}, "11-06": {"max_temp": "12", "min_temp": "0", "weather": "dy", "wind": "12"}, "11-07": {"max_temp": "13", "min_temp": "3", "weather": "dy", "wind": "12"}, "11-08": {"max_temp": "14", "min_temp": "2", "weather": "dy", "wind": "12"}, "11-09": {"max_temp": "15", "min_temp": "1", "weather": "qdy", "wind": "34"}, "11-10": {"max_temp": "11", "min_temp": "0", "weather": "dy", "wind": "12"}, "11-11": {"max_temp": "13", "min_temp": "1", "weather": "dyq", "wind": "12"}, "11-12": {"max_temp": "14", "min_temp": "2", "weather": "q", "wind": "12"}, "11-13": {"max_temp": "13", "min_temp": "5", "weather": "dy", "wind": "12"}, "11-14": {"max_temp": "13", "min_temp": "5", "weather": "dy", "wind": "12"}, "11-15": {"max_temp": "8", "min_temp": "1", "weather": "xydy", "wind": "34"}, "11-16": {"max_temp": "8", "min_temp": "-1", "weather": "q", "wind": "12"}, "11-17": {"max_temp": "9", "min_temp": "-2", "weather": "dyq", "wind": "12"}, "11-18": {"max_temp": "11", "min_temp": "-3", "weather": "q", "wind": "34"}, "11-19": {"max_temp": "10", "min_temp": "-2", "weather": "qdy", "wind": "12"}, "11-20": {"max_temp": "9", "min_temp": "-1", "weather": "dy", "wind": "12"}, "11-21": {"max_temp": "9", "min_temp": "-3", "weather": "q", "wind": "2"}, "11-22": {"max_temp": "8", "min_temp": "-3", "weather": "qdy", "wind": "1"}, "11-23": {"max_temp": "7", "min_temp": "0", "weather": "dy", "wind": "2"}, "11-24": {"max_temp": "9", "min_temp": "-3", "weather": "qdy", "wind": "2"}, "11-25": {"max_temp": "10", "min_temp": "-3", "weather": "q", "wind": "1"}, "11-26": {"max_temp": "10", "min_temp": "0", "weather": "dy", "wind": "1"}, "11-27": {"max_temp": "9", "min_temp": "-3", "weather": "qdy", "wind": "2"}, "11-28": {"max_temp": "8", "min_temp": "-3", "weather": "q", "wind": "1"}, "11-29": {"max_temp": "7", "min_temp": "-4", "weather": "q", "wind": "1"}, "11-30": {"max_temp": "8", "min_temp": "-3", "weather": "q", "wind": "1"}, "12-01": {"max_temp": "7", "min_temp": "0", "weather": "dy", "wind": "1"}, "12-02": {"max_temp": "9", "min_temp": "2", "weather": "dy", "wind": "1"}, "12-03": {"max_temp": "8", "min_temp": "-3", "weather": "dyq", "wind": "3"}, "12-04": {"max_temp": "4", "min_temp": "-6", "weather": "qdy", "wind": "2"}, "12-05": {"max_temp": "1", "min_temp": "-4", "weather": "dy", "wind": "1"}, "12-06": {"max_temp": "-2", "min_temp": "-9", "weather": "q", "wind": "3"}, "12-07": {"max_temp": "-4", "min_temp": "-10", "weather": "q", "wind": "3"}, "12-08": {"max_temp": "-2", "min_temp": "-10", "weather": "qdy", "wind": "2"}, "12-09": {"max_temp": "-1", "min_temp": "-10", "weather": "dyq", "wind": "1"}}
\ No newline at end of file
PaddleRec/ctr/dcn/cluster_train.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
import argparse
import os
import sys
import time
from collections import OrderedDict
import paddle.fluid as fluid
from network import DCN
import utils
def boolean_string(s):
if s.lower() not in {'false', 'true'}:
raise ValueError('Not a valid boolean string')
return s.lower() == 'true'
def parse_args():
parser = argparse.ArgumentParser("dcn cluster train.")
parser.add_argument(
'--train_data_dir',
type=str,
default='dist_data/dist_train_data',
help='The path of train data')
parser.add_argument(
'--test_valid_data_dir',
type=str,
default='dist_data/dist_test_valid_data',
help='The path of test and valid data')
parser.add_argument(
'--vocab_dir',
type=str,
default='dist_data/vocab',
help='The path of generated vocabs')
parser.add_argument(
'--cat_feat_num',
type=str,
default='dist_data/cat_feature_num.txt',
help='The path of generated cat_feature_num.txt')
parser.add_argument(
'--batch_size', type=int, default=512, help="Batch size")
parser.add_argument('--num_epoch', type=int, default=10, help="train epoch")
parser.add_argument(
'--model_output_dir',
type=str,
default='models',
help='The path for model to store')
parser.add_argument(
'--num_thread', type=int, default=1, help='The number of threads')
parser.add_argument('--test_epoch', type=str, default='1')
parser.add_argument(
'--dnn_hidden_units',
nargs='+',
type=int,
default=[1024, 1024],
help='DNN layers and hidden units')
parser.add_argument(
'--cross_num',
type=int,
default=6,
help='The number of Cross network layers')
parser.add_argument('--lr', type=float, default=1e-4, help='Learning rate')
parser.add_argument(
'--l2_reg_cross',
type=float,
default=1e-5,
help='Cross net l2 regularizer coefficient')
parser.add_argument(
'--use_bn',
type=boolean_string,
default=True,
help='Whether use batch norm in dnn part')
parser.add_argument(
'--is_sparse',
action='store_true',
required=False,
default=False,
help='embedding will use sparse or not, (default: False)')
parser.add_argument(
'--clip_by_norm', type=float, default=100.0, help="gradient clip norm")
parser.add_argument('--print_steps', type=int, default=5)
parser.add_argument('--use_gpu', type=int, default=1)
# dist params
parser.add_argument('--is_local', type=int, default=1, help='whether local')
parser.add_argument(
'--num_devices', type=int, default=1, help='Number of GPU devices')
parser.add_argument(
'--role', type=str, default='pserver', help='trainer or pserver')
parser.add_argument(
'--endpoints',
type=str,
default='127.0.0.1:6000',
help='The pserver endpoints, like: 127.0.0.1:6000, 127.0.0.1:6001')
parser.add_argument(
'--current_endpoint',
type=str,
default='127.0.0.1:6000',
help='The current_endpoint')
parser.add_argument(
'--trainer_id',
type=int,
default=0,
help='trainer id ,only trainer_id=0 save model')
parser.add_argument(
'--trainers',
type=int,
default=1,
help='The num of trianers, (default: 1)')
args = parser.parse_args()
return args
def train():
""" do training """
args = parse_args()
print(args)
if args.trainer_id == 0 and not os.path.isdir(args.model_output_dir):
os.mkdir(args.model_output_dir)
cat_feat_dims_dict = OrderedDict()
for line in open(args.cat_feat_num):
spls = line.strip().split()
assert len(spls) == 2
cat_feat_dims_dict[spls[0]] = int(spls[1])
dcn_model = DCN(args.cross_num, args.dnn_hidden_units, args.l2_reg_cross,
args.use_bn, args.clip_by_norm, cat_feat_dims_dict,
args.is_sparse)
dcn_model.build_network()
optimizer = fluid.optimizer.Adam(learning_rate=args.lr)
optimizer.minimize(dcn_model.loss)
def train_loop(main_program):
""" train network """
start_time = time.time()
dataset = fluid.DatasetFactory().create_dataset()
dataset.set_use_var(dcn_model.data_list)
pipe_command = 'python reader.py {}'.format(args.vocab_dir)
dataset.set_pipe_command(pipe_command)
dataset.set_batch_size(args.batch_size)
dataset.set_thread(args.num_thread)
train_filelist = [
os.path.join(args.train_data_dir, fname)
for fname in next(os.walk(args.train_data_dir))[2]
]
dataset.set_filelist(train_filelist)
if args.use_gpu == 1:
exe = fluid.Executor(fluid.CUDAPlace(0))
dataset.set_thread(1)
else:
exe = fluid.Executor(fluid.CPUPlace())
dataset.set_thread(args.num_thread)
exe.run(fluid.default_startup_program())
for epoch_id in range(args.num_epoch):
start = time.time()
sys.stderr.write('\nepoch%d start ...\n' % (epoch_id + 1))
exe.train_from_dataset(
program=main_program,
dataset=dataset,
fetch_list=[
dcn_model.loss, dcn_model.avg_logloss, dcn_model.auc_var
],
fetch_info=['total_loss', 'avg_logloss', 'auc'],
debug=False,
print_period=args.print_steps)
model_dir = os.path.join(args.model_output_dir,
'epoch_' + str(epoch_id + 1), "checkpoint")
sys.stderr.write('epoch%d is finished and takes %f s\n' % (
(epoch_id + 1), time.time() - start))
if args.trainer_id == 0: # only trainer 0 save model
print("save model in {}".format(model_dir))
fluid.save(main_program, model_dir)
print("train time cost {:.4f}".format(time.time() - start_time))
print("finish training")
if args.is_local:
print("run local training")
train_loop(fluid.default_main_program())
else:
print("run distribute training")
t = fluid.DistributeTranspiler()
t.transpile(
args.trainer_id, pservers=args.endpoints, trainers=args.trainers)
if args.role == "pserver":
print("run psever")
pserver_prog, pserver_startup = t.get_pserver_programs(
args.current_endpoint)
exe = fluid.Executor(fluid.CPUPlace())
exe.run(pserver_startup)
exe.run(pserver_prog)
elif args.role == "trainer":
print("run trainer")
train_loop(t.get_trainer_program())
if __name__ == "__main__":
utils.check_version()
train()
PaddleRec/ctr/dcn/cluster_train.sh 已删除 100755 → 0
浏览文件 @ 57c0bf10
#!/bin/bash
#export GLOG_v=30
#export GLOG_logtostderr=1
# start pserver0
python -u cluster_train.py \
--train_data_dir dist_data/dist_train_data \
--model_output_dir cluster_model \
--is_local 0 \
--is_sparse \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6000 \
--trainers 2 \
> pserver0.log 2>&1 &
# start pserver1
python -u cluster_train.py \
--train_data_dir dist_data/dist_train_data \
--model_output_dir cluster_model \
--is_local 0 \
--is_sparse \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6001 \
--trainers 2 \
> pserver1.log 2>&1 &
# start trainer0
#CUDA_VISIBLE_DEVICES=1 python cluster_train.py \
python -u cluster_train.py \
--train_data_dir dist_data/dist_train_data \
--model_output_dir cluster_model \
--use_gpu 0 \
--is_local 0 \
--is_sparse \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 0 \
> trainer0.log 2>&1 &
# start trainer1
#CUDA_VISIBLE_DEVICES=2 python cluster_train.py \
python -u cluster_train.py \
--train_data_dir dist_data/dist_train_data \
--model_output_dir cluster_model \
--use_gpu 0 \
--is_local 0 \
--is_sparse \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 1 \
> trainer1.log 2>&1 &
echo "2 pservers and 2 trainers started."
\ No newline at end of file
PaddleRec/ctr/dcn/reader.py
浏览文件 @ a8a58316
......@@ -72,8 +72,8 @@ class CriteoDataset(dg.MultiSlotDataGenerator):
yield label_feat_list
import paddle
batch_iter = paddle.batch(
paddle.reader.buffered(
batch_iter = fluid.io.batch(
fluid.io.buffered(
local_iter, size=buf_size), batch_size=batch)
return batch_iter
......
PaddleRec/ctr/deepfm/cluster_train.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
import argparse
import os
import sys
import time
from network_conf import ctr_deepfm_model
import paddle.fluid as fluid
import utils
def parse_args():
parser = argparse.ArgumentParser("deepfm cluster train.")
parser.add_argument(
'--train_data_dir',
type=str,
default='dist_data/dist_train_data',
help='The path of train data (default: data/train_data)')
parser.add_argument(
'--test_data_dir',
type=str,
default='dist_data/dist_test_data',
help='The path of test data (default: models)')
parser.add_argument(
'--feat_dict',
type=str,
default='dist_data/aid_data/feat_dict_10.pkl2',
help='The path of feat_dict')
parser.add_argument(
'--batch_size',
type=int,
default=100,
help="The size of mini-batch (default:100)")
parser.add_argument(
'--embedding_size',
type=int,
default=10,
help="The size for embedding layer (default:10)")
parser.add_argument(
'--num_epoch',
type=int,
default=10,
help="The number of epochs to train (default: 50)")
parser.add_argument(
'--model_output_dir',
type=str,
required=True,
help='The path for model to store (default: models)')
parser.add_argument(
'--num_thread',
type=int,
default=1,
help='The number of threads (default: 1)')
parser.add_argument('--test_epoch', type=str, default='1')
parser.add_argument(
'--layer_sizes',
nargs='+',
type=int,
default=[400, 400, 400],
help='The size of each layers (default: [10, 10, 10])')
parser.add_argument(
'--act',
type=str,
default='relu',
help='The activation of each layers (default: relu)')
parser.add_argument(
'--is_sparse',
action='store_true',
required=False,
default=False,
help='embedding will use sparse or not, (default: False)')
parser.add_argument(
'--lr', type=float, default=1e-4, help='Learning rate (default: 1e-4)')
parser.add_argument(
'--reg', type=float, default=1e-4, help=' (default: 1e-4)')
parser.add_argument('--num_field', type=int, default=39)
parser.add_argument('--num_feat', type=int, default=141443)
parser.add_argument('--use_gpu', type=int, default=1)
# dist params
parser.add_argument('--is_local', type=int, default=1, help='whether local')
parser.add_argument(
'--num_devices', type=int, default=1, help='Number of GPU devices')
parser.add_argument(
'--role', type=str, default='pserver', help='trainer or pserver')
parser.add_argument(
'--endpoints',
type=str,
default='127.0.0.1:6000',
help='The pserver endpoints, like: 127.0.0.1:6000, 127.0.0.1:6001')
parser.add_argument(
'--current_endpoint',
type=str,
default='127.0.0.1:6000',
help='The current_endpoint')
parser.add_argument(
'--trainer_id',
type=int,
default=0,
help='trainer id ,only trainer_id=0 save model')
parser.add_argument(
'--trainers',
type=int,
default=1,
help='The num of trianers, (default: 1)')
args = parser.parse_args()
return args
def train():
""" do training """
args = parse_args()
print(args)
if args.trainer_id == 0 and not os.path.isdir(args.model_output_dir):
os.mkdir(args.model_output_dir)
loss, auc, data_list, auc_states = ctr_deepfm_model(
args.embedding_size, args.num_field, args.num_feat, args.layer_sizes,
args.act, args.reg, args.is_sparse)
optimizer = fluid.optimizer.SGD(
learning_rate=args.lr,
regularization=fluid.regularizer.L2DecayRegularizer(args.reg))
optimizer.minimize(loss)
def train_loop(main_program):
""" train network """
start_time = time.time()
dataset = fluid.DatasetFactory().create_dataset()
dataset.set_use_var(data_list)
pipe_command = 'python criteo_reader.py {}'.format(args.feat_dict)
dataset.set_pipe_command(pipe_command)
dataset.set_batch_size(args.batch_size)
dataset.set_thread(args.num_thread)
train_filelist = [
os.path.join(args.train_data_dir, x)
for x in os.listdir(args.train_data_dir)
]
if args.use_gpu == 1:
exe = fluid.Executor(fluid.CUDAPlace(0))
dataset.set_thread(1)
else:
exe = fluid.Executor(fluid.CPUPlace())
dataset.set_thread(args.num_thread)
exe.run(fluid.default_startup_program())
for epoch_id in range(args.num_epoch):
start = time.time()
sys.stderr.write('\nepoch%d start ...\n' % (epoch_id + 1))
dataset.set_filelist(train_filelist)
exe.train_from_dataset(
program=main_program,
dataset=dataset,
fetch_list=[loss, auc],
fetch_info=['epoch %d batch loss' % (epoch_id + 1), "auc"],
print_period=5,
debug=False)
model_dir = os.path.join(args.model_output_dir,
'epoch_' + str(epoch_id + 1))
sys.stderr.write('epoch%d is finished and takes %f s\n' % (
(epoch_id + 1), time.time() - start))
if args.trainer_id == 0: # only trainer 0 save model
print("save model in {}".format(model_dir))
fluid.save(main_program, model_dir)
print("train time cost {:.4f}".format(time.time() - start_time))
print("finish training")
if args.is_local:
print("run local training")
train_loop(fluid.default_main_program())
else:
print("run distribute training")
t = fluid.DistributeTranspiler()
t.transpile(
args.trainer_id, pservers=args.endpoints, trainers=args.trainers)
if args.role == "pserver":
print("run psever")
pserver_prog, pserver_startup = t.get_pserver_programs(
args.current_endpoint)
exe = fluid.Executor(fluid.CPUPlace())
exe.run(pserver_startup)
exe.run(pserver_prog)
elif args.role == "trainer":
print("run trainer")
train_loop(t.get_trainer_program())
if __name__ == "__main__":
utils.check_version()
train()
PaddleRec/ctr/deepfm/cluster_train.sh 已删除 100755 → 0
浏览文件 @ 57c0bf10
#!/bin/bash
#export GLOG_v=30
#export GLOG_logtostderr=1
# start pserver0
python -u cluster_train.py \
--train_data_dir dist_data/dist_train_data \
--model_output_dir cluster_model \
--is_local 0 \
--is_sparse \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6000 \
--trainers 2 \
> pserver0.log 2>&1 &
# start pserver1
python -u cluster_train.py \
--train_data_dir dist_data/dist_train_data \
--model_output_dir cluster_model \
--is_local 0 \
--is_sparse \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6001 \
--trainers 2 \
> pserver1.log 2>&1 &
# start trainer0
#CUDA_VISIBLE_DEVICES=1 python cluster_train.py \
python -u cluster_train.py \
--train_data_dir dist_data/dist_train_data \
--model_output_dir cluster_model \
--use_gpu 0 \
--is_local 0 \
--is_sparse \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 0 \
> trainer0.log 2>&1 &
# start trainer1
#CUDA_VISIBLE_DEVICES=2 python cluster_train.py \
python -u cluster_train.py \
--train_data_dir dist_data/dist_train_data \
--model_output_dir cluster_model \
--use_gpu 0 \
--is_local 0 \
--is_sparse \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 1 \
> trainer1.log 2>&1 &
echo "2 pservers and 2 trainers started."
\ No newline at end of file
PaddleRec/ctr/deepfm/infer.py
浏览文件 @ a8a58316
......@@ -30,7 +30,7 @@ def infer():
]
criteo_dataset = CriteoDataset()
criteo_dataset.setup(args.feat_dict)
test_reader = paddle.batch(
test_reader = fluid.io.batch(
criteo_dataset.test(test_files), batch_size=args.batch_size)
startup_program = fluid.framework.Program()
......
PaddleRec/ctr/deepfm_dygraph/data_reader.py
浏览文件 @ a8a58316
......@@ -6,6 +6,7 @@ import pickle
import random
import paddle
import paddle.fluid as fluid
class DataGenerator(object):
......@@ -58,7 +59,7 @@ class DataGenerator(object):
if not cycle:
break
return paddle.batch(_reader, batch_size=batch_size)
return fluid.io.batch(_reader, batch_size=batch_size)
def data_reader(batch_size,
......
PaddleRec/ctr/din/README.md
浏览文件 @ a8a58316
......@@ -8,8 +8,6 @@
├── train.py # 训练脚本
├── infer.py # 预测脚本
├── network.py # 网络结构
├── cluster_train.py # 多机训练
├── cluster_train.sh # 多机训练脚本
├── reader.py # 和读取数据相关的函数
├── data/
├── build_dataset.py # 文本数据转化为paddle数据
......@@ -129,12 +127,3 @@ CUDA_VISIBLE_DEVICES=3 python infer.py --model_path 'din_amazon/global_step_4000
```text
2019-02-22 11:22:58,804 - INFO - TEST --> loss: [0.47005194] auc:0.863794952818
```
## 多机训练
可参考cluster_train.py 配置多机环境
运行命令本地模拟多机场景
```
sh cluster_train.sh
```
PaddleRec/ctr/din/cluster_train.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
import sys
import logging
import time
import numpy as np
import argparse
import paddle.fluid as fluid
import paddle
import time
import network
import reader
import random
logging.basicConfig(format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger("fluid")
logger.setLevel(logging.INFO)
def parse_args():
parser = argparse.ArgumentParser("din")
parser.add_argument(
'--config_path',
type=str,
default='data/config.txt',
help='dir of config')
parser.add_argument(
'--train_dir',
type=str,
default='data/paddle_train.txt',
help='dir of train file')
parser.add_argument(
'--model_dir',
type=str,
default='din_amazon/',
help='dir of saved model')
parser.add_argument(
'--batch_size', type=int, default=16, help='number of batch size')
parser.add_argument(
'--epoch_num', type=int, default=200, help='number of epoch')
parser.add_argument(
'--use_cuda', type=int, default=0, help='whether to use gpu')
parser.add_argument(
'--parallel',
type=int,
default=0,
help='whether to use parallel executor')
parser.add_argument(
'--base_lr', type=float, default=0.85, help='based learning rate')
parser.add_argument(
'--role', type=str, default='pserver', help='trainer or pserver')
parser.add_argument(
'--endpoints',
type=str,
default='127.0.0.1:6000',
help='The pserver endpoints, like: 127.0.0.1:6000, 127.0.0.1:6001')
parser.add_argument(
'--current_endpoint',
type=str,
default='127.0.0.1:6000',
help='The current_endpoint')
parser.add_argument(
'--trainer_id',
type=int,
default=0,
help='trainer id ,only trainer_id=0 save model')
parser.add_argument(
'--trainers',
type=int,
default=1,
help='The num of trianers, (default: 1)')
args = parser.parse_args()
return args
def train():
args = parse_args()
config_path = args.config_path
train_path = args.train_dir
epoch_num = args.epoch_num
use_cuda = True if args.use_cuda else False
use_parallel = True if args.parallel else False
logger.info("reading data begins")
user_count, item_count, cat_count = reader.config_read(config_path)
#data_reader, max_len = reader.prepare_reader(train_path, args.batch_size)
logger.info("reading data completes")
avg_cost, pred = network.network(item_count, cat_count, 433)
base_lr = args.base_lr
boundaries = [410000]
values = [base_lr, 0.2]
sgd_optimizer = fluid.optimizer.SGD(
learning_rate=fluid.layers.piecewise_decay(
boundaries=boundaries, values=values))
sgd_optimizer.minimize(avg_cost)
def train_loop(main_program):
data_reader, max_len = reader.prepare_reader(train_path,
args.batch_size)
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
exe = fluid.Executor(place)
exe.run(fluid.default_startup_program())
feed_list = [
"hist_item_seq", "hist_cat_seq", "target_item", "target_cat",
"label", "mask", "target_item_seq", "target_cat_seq"
]
loader = fluid.io.DataLoader.from_generator(
feed_list=feed_list, capacity=10000, iterable=True)
loader.set_sample_list_generator(data_reader, places=place)
if use_parallel:
train_exe = fluid.ParallelExecutor(
use_cuda=use_cuda,
loss_name=avg_cost.name,
main_program=main_program)
else:
train_exe = exe
logger.info("train begins")
global_step = 0
PRINT_STEP = 1000
start_time = time.time()
loss_sum = 0.0
for id in range(epoch_num):
epoch = id + 1
for data in loader():
global_step += 1
results = train_exe.run(main_program,
feed=data,
fetch_list=[avg_cost.name, pred.name],
return_numpy=True)
loss_sum += results[0].mean()
if global_step % PRINT_STEP == 0:
logger.info(
"epoch: %d\tglobal_step: %d\ttrain_loss: %.4f\t\ttime: %.2f"
% (epoch, global_step, loss_sum / PRINT_STEP,
time.time() - start_time))
start_time = time.time()
loss_sum = 0.0
if (global_step > 400000 and
global_step % PRINT_STEP == 0) or (
global_step < 400000 and
global_step % 50000 == 0):
save_dir = args.model_dir + "/global_step_" + str(
global_step)
feed_var_name = [
"hist_item_seq", "hist_cat_seq", "target_item",
"target_cat", "label", "mask", "target_item_seq",
"target_cat_seq"
]
fetch_vars = [avg_cost, pred]
fluid.io.save_inference_model(save_dir, feed_var_name,
fetch_vars, exe)
train_exe.close()
t = fluid.DistributeTranspiler()
t.transpile(
args.trainer_id, pservers=args.endpoints, trainers=args.trainers)
if args.role == "pserver":
logger.info("run psever")
prog, startup = t.get_pserver_programs(args.current_endpoint)
exe = fluid.Executor(fluid.CPUPlace())
exe.run(startup)
exe.run(prog)
elif args.role == "trainer":
logger.info("run trainer")
train_loop(t.get_trainer_program())
if __name__ == "__main__":
train()
PaddleRec/ctr/din/cluster_train.sh 已删除 100644 → 0
浏览文件 @ 57c0bf10
#!/bin/bash
#export GLOG_v=30
#export GLOG_logtostderr=1
python -u cluster_train.py \
--config_path 'data/config.txt' \
--train_dir 'data/paddle_train.txt' \
--batch_size 32 \
--epoch_num 100 \
--use_cuda 0 \
--parallel 0 \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6000 \
--trainers 2 \
> pserver0.log 2>&1 &
python -u cluster_train.py \
--config_path 'data/config.txt' \
--train_dir 'data/paddle_train.txt' \
--batch_size 32 \
--epoch_num 100 \
--use_cuda 0 \
--parallel 0 \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6001 \
--trainers 2 \
> pserver1.log 2>&1 &
python -u cluster_train.py \
--config_path 'data/config.txt' \
--train_dir 'data/paddle_train.txt' \
--batch_size 32 \
--epoch_num 100 \
--use_cuda 0 \
--parallel 0 \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 0 \
> trainer0.log 2>&1 &
python -u cluster_train.py \
--config_path 'data/config.txt' \
--train_dir 'data/paddle_train.txt' \
--batch_size 32 \
--epoch_num 100 \
--use_cuda 0 \
--parallel 0 \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 1 \
> trainer1.log 2>&1 &
PaddleRec/ctr/dnn/README.md
浏览文件 @ a8a58316
......@@ -154,7 +154,7 @@ def embedding_layer(input):
return fluid.layers.embedding(
input=input,
is_sparse=True,
size=[args.sparse_feature_dim,
size=[args.sparse_feature_dim,
args.embedding_size],
param_attr=fluid.ParamAttr(
name="SparseFeatFactors",
......@@ -168,7 +168,7 @@ sparse_embed_seq = list(map(embedding_layer, inputs[1:-1])) # [C1~C26]
将离散数据通过embedding查表得到的值，与连续数据的输入进行`concat`操作，合为一个整体输入，作为全链接层的原始输入。我们共设计了3层FC，每层FC的输出维度都为400，每层FC都后接一个`relu`激活函数，每层FC的初始化方式为符合正态分布的随机初始化，标准差与上一层的输出维度的平方根成反比。
```python
concated = fluid.layers.concat(sparse_embed_seq + inputs[0:1], axis=1)
fc1 = fluid.layers.fc(
input=concated,
size=400,
......@@ -271,9 +271,9 @@ continuous_range_ = range(1, 14)
categorical_range_ = range(14, 40)
class CriteoDataset(dg.MultiSlotDataGenerator):
def generate_sample(self, line):
def reader():
features = line.rstrip('\n').split('\t')
dense_feature = []
......@@ -363,12 +363,12 @@ fleet.init(role) #必不可少的步骤，初始化节点！
> PaddleCloudRoleMaker()是怎样判断当前节点所扮演的角色的？
>
>
> Paddle参数服务器模式中，使用各个节点机器的环境变量来确定当前节点的角色。为了能准确无误的分配角色，在每个节点上，我们都需要指定如下环境变量：
> #### 共有的环境变量
> - export PADDLE_TRAINERS_NUM=2 # 训练节点数
> - export PADDLE_PSERVERS_IP_PORT_LIST="127.0.0.1:36011,127.0.0.1:36012" # 各个pserver的ip:port 组合构成的字符串
>
>
> #### Pserver特有的环境变量
> - export TRAINING_ROLE=PSERVER # 当前节点的角色是PSERVER
> - export PADDLE_PORT=36011 # 当前PSERVER的通信端口
......@@ -376,7 +376,7 @@ fleet.init(role) #必不可少的步骤，初始化节点！
> #### Trainer特有的环境变量
> - export TRAINING_ROLE=TRAINER # 当前节点的角色是TRAINER
> - export PADDLE_TRAINER_ID=0 # 当前Trainer节点的编号,范围为[0，PADDLE_TRAINERS_NUM)
>
>
> 完成上述环境变量指定后，`PaddleCloudRoleMaker()`便可以正常的运行，决定当前节点的角色。
......@@ -388,7 +388,7 @@ Paddle的`参数服务器`模式分布式训练有很多种类型，根据通信
ctr_model = CTR()
inputs = ctr_model.input_data(args)
avg_cost, auc_var, batch_auc_var = ctr_model.net(inputs,args)
# 选择反向更新优化策略
optimizer = fluid.optimizer.Adam(args.learning_rate)
optimizer.minimize(avg_cost)
......@@ -431,7 +431,7 @@ if fleet.is_server():
fleet.run_server()
```
- 启动Worker
启动训练节点，训练节点首先调用`init_worker()`来完成节点初始化，然后执行`fleet.startup_program`，从服务器端同步参数的初始化值。接着，和本地训练完全一致，通过执行`fleet.main_program`来完成整个训练过程，并保存模型。最后调用`fleet.stop_worker()`关闭训练节点。
```python
elif fleet.is_worker():
......@@ -441,7 +441,7 @@ elif fleet.is_worker():
# 初始化含有分布式流程的fleet.startup_program
exe.run(fleet.startup_program))
# 引入数据读取dataset
dataset = get_dataset(inputs,params)
......@@ -458,10 +458,10 @@ elif fleet.is_worker():
# 默认使用0号节点保存模型
if params.test and fleet.is_first_worker():
model_path = (str(params.model_path) + "/"+"epoch_" + str(epoch))
fluid.io.save_persistables(executor=exe, dirname=model_path)
fleet.save_persistables(executor=exe, dirname=model_path)
# 训练结束，调用stop_worker()通知pserver
fleet.stop_worker()
fleet.stop_worker()
logger.info("Distribute Train Success!")
return train_result
```
......@@ -504,7 +504,7 @@ sh local_cluster.sh
便可以开启分布式模拟训练，默认启用2x2的训练模式。Trainer与Pserver的运行日志，存放于`./log/`文件夹，保存的模型位于`./models/`，使用默认配置运行后，理想输出为：
- pserver.0.log
```bash
get_pserver_program() is deprecated, call get_pserver_programs() to get pserver main and startup in a single call.
I1126 07:37:49.952580 15056 grpc_server.cc:477] Server listening on 127.0.0.1:36011 successful, selected port: 36011
```
......@@ -558,9 +558,9 @@ I1126 07:38:28.947571 14715 communicator.cc:363] Communicator stop done
2. 在很多应用场景中，分布式训练出的模型与实际上线的模型不一致，仅使用分布式训练出的参数值，参与其他网络的预测，在这样的场景中，就更无必要保存模型结构了。
> 什么是长期变量？
>
>
> 在Paddle Fluid中，模型变量可以分为以下几种类型：
>
>
> 1. 模型参数：是深度学习模型中被训练和学习的量。由`fluid.framwork.Parameter()`产生，是`fluid.framework.Variable()`的派生类。
> 2. 长期变量 ：是在整个训练过程中持续存在，不会因为一个迭代结束而销毁的变量，所有的模型参数都是长期变量，但并非所有的长期变量都是模型参数。长期变量通过将`fluid.framework.Varibale()`中的`psersistable`属性设置为`True`来声明。长期变量是模型的核心参数。
> 3. 临时变量：不属于上述两种类别的所有变量都是临时变量，只在一个训练迭代中存在，在每一个迭代结束后，所有的临时变量都会被销毁，然后在下一个迭代开始时，创建新的临时变量。例如输入的训练数据，中间层layer的输出等等。
......@@ -632,7 +632,7 @@ with fluid.framework.program_guard(test_program, startup_program):
```
这是容易理解的，因为在测试时，我们要从零开始，保证预测program的干净，没有其他的影响因素。
- 在创建预测网络时，我们加入了`with fluid.unique_name.guard():`，它的作用是让所有新建的参数的自动编号再次从零开始。Paddle的参数`Variable`以变量名作为区分手段，保证变量名相同，就可以从保存的模型中找到对应参数。
paddle创建的临时变量，编号会自动顺延，如果没有指定变量名，可以观察到这一现象，比如：`fc_1.w_0`->`fc_2.w_0`，想要共享相同的参数，必需要保证编号可以对应。
### 测试数据的读取
......@@ -774,14 +774,14 @@ python -u train.py --is_cloud=1
运行该命令时，若pserver还未就绪，可在日志输出中看到如下信息：
> server not ready, wait 3 sec to retry...
>
>
> not ready endpoints:['10.89.176.11:36000', '10.89.176.12:36000']
worker进程将持续等待，直到server开始监听，或等待超时。
当pserver都准备就绪后，可以在日志输出看到如下信息：
> I0317 11:38:48.099179 16719 communicator.cc:271] Communicator start
>
>
> I0317 11:38:49.838711 16719 rpc_client.h:107] init rpc client with trainer_id 0
至此，分布式训练启动完毕，将开始训练，祝您好运。
PaddleRec/ctr/dnn/infer.py
浏览文件 @ a8a58316
......@@ -30,8 +30,7 @@ logger.setLevel(logging.INFO)
def parse_args():
parser = argparse.ArgumentParser(
description="PaddlePaddle CTR-DNN example")
parser = argparse.ArgumentParser(description="PaddlePaddle CTR-DNN example")
# -------------Data & Model Path-------------
parser.add_argument(
'--test_files_path',
......@@ -54,8 +53,7 @@ def parse_args():
'--infer_epoch',
type=int,
default=0,
help='Specify which epoch to run infer'
)
help='Specify which epoch to run infer')
# -------------Network parameter-------------
parser.add_argument(
'--embedding_size',
......@@ -68,10 +66,7 @@ def parse_args():
default=1000001,
help='sparse feature hashing space for index processing')
parser.add_argument(
'--dense_feature_dim',
type=int,
default=13,
help='dense feature shape')
'--dense_feature_dim', type=int, default=13, help='dense feature shape')
# -------------device parameter-------------
parser.add_argument(
......@@ -102,10 +97,11 @@ def run_infer(args, model_path):
place = fluid.CPUPlace()
train_generator = generator.CriteoDataset(args.sparse_feature_dim)
file_list = [
os.path.join(args.test_files_path, x) for x in os.listdir(args.test_files_path)
os.path.join(args.test_files_path, x)
for x in os.listdir(args.test_files_path)
]
test_reader = paddle.batch(train_generator.test(file_list),
batch_size=args.batch_size)
test_reader = fluid.io.batch(
train_generator.test(file_list), batch_size=args.batch_size)
startup_program = fluid.framework.Program()
test_program = fluid.framework.Program()
ctr_model = CTR()
......@@ -171,13 +167,15 @@ if __name__ == "__main__":
model_list = []
for _, dir, _ in os.walk(args.model_path):
for model in dir:
if "epoch" in model and args.infer_epoch == int(model.split('_')[-1]):
if "epoch" in model and args.infer_epoch == int(
model.split('_')[-1]):
path = os.path.join(args.model_path, model)
model_list.append(path)
if len(model_list) == 0:
logger.info("There is no satisfactory model {} at path {}, please check your start command & env. ".format(
str("epoch_")+str(args.infer_epoch), args.model_path))
logger.info(
"There is no satisfactory model {} at path {}, please check your start command & env. ".
format(str("epoch_") + str(args.infer_epoch), args.model_path))
for model in model_list:
logger.info("Test model {}".format(model))
......
PaddleRec/ctr/xdeepfm/cluster_train.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
import argparse
import os
import sys
import time
import network_conf
import paddle.fluid as fluid
import utils
def parse_args():
parser = argparse.ArgumentParser("xdeepfm cluster train.")
parser.add_argument(
'--train_data_dir',
type=str,
default='data/train_data',
help='The path of train data (default: data/train_data)')
parser.add_argument(
'--test_data_dir',
type=str,
default='data/test_data',
help='The path of test data (default: models)')
parser.add_argument(
'--batch_size',
type=int,
default=100,
help="The size of mini-batch (default:100)")
parser.add_argument(
'--embedding_size',
type=int,
default=10,
help="The size for embedding layer (default:10)")
parser.add_argument(
'--num_epoch',
type=int,
default=10,
help="The number of epochs to train (default: 10)")
parser.add_argument(
'--model_output_dir',
type=str,
required=True,
help='The path for model to store (default: models)')
parser.add_argument(
'--num_thread',
type=int,
default=1,
help='The number of threads (default: 1)')
parser.add_argument('--test_epoch', type=str, default='1')
parser.add_argument(
'--layer_sizes_dnn',
nargs='+',
type=int,
default=[10, 10, 10],
help='The size of each layers')
parser.add_argument(
'--layer_sizes_cin',
nargs='+',
type=int,
default=[10, 10],
help='The size of each layers')
parser.add_argument(
'--act',
type=str,
default='relu',
help='The activation of each layers (default: relu)')
parser.add_argument(
'--lr', type=float, default=1e-1, help='Learning rate (default: 1e-4)')
parser.add_argument(
'--reg', type=float, default=1e-4, help=' (default: 1e-4)')
parser.add_argument('--num_field', type=int, default=39)
parser.add_argument('--num_feat', type=int, default=28651)
parser.add_argument(
'--model_name',
type=str,
default='ctr_xdeepfm_model',
help='The name of model (default: ctr_xdeepfm_model)')
parser.add_argument('--use_gpu', type=int, default=1)
parser.add_argument('--print_steps', type=int, default=50)
parser.add_argument('--is_local', type=int, default=1, help='whether local')
parser.add_argument(
'--is_sparse',
action='store_true',
required=False,
default=False,
help='embedding will use sparse or not, (default: False)')
# dist params
parser.add_argument(
'--num_devices', type=int, default=1, help='Number of GPU devices')
parser.add_argument(
'--role', type=str, default='pserver', help='trainer or pserver')
parser.add_argument(
'--endpoints',
type=str,
default='127.0.0.1:6000',
help='The pserver endpoints, like: 127.0.0.1:6000, 127.0.0.1:6001')
parser.add_argument(
'--current_endpoint',
type=str,
default='127.0.0.1:6000',
help='The current_endpoint')
parser.add_argument(
'--trainer_id',
type=int,
default=0,
help='trainer id ,only trainer_id=0 save model')
parser.add_argument(
'--trainers',
type=int,
default=1,
help='The num of trianers, (default: 1)')
args = parser.parse_args()
return args
def train():
""" do training """
args = parse_args()
print(args)
if not os.path.isdir(args.model_output_dir):
os.mkdir(args.model_output_dir)
loss, auc, data_list, auc_states = eval('network_conf.' + args.model_name)(
args.embedding_size, args.num_field, args.num_feat,
args.layer_sizes_dnn, args.act, args.reg, args.layer_sizes_cin,
args.is_sparse)
optimizer = fluid.optimizer.SGD(
learning_rate=args.lr,
regularization=fluid.regularizer.L2DecayRegularizer(args.reg))
optimizer.minimize(loss)
def train_loop(main_program):
""" train network """
start_time = time.time()
dataset = fluid.DatasetFactory().create_dataset()
dataset.set_use_var(data_list)
dataset.set_pipe_command('python criteo_reader.py')
dataset.set_batch_size(args.batch_size)
dataset.set_filelist([
os.path.join(args.train_data_dir, x)
for x in os.listdir(args.train_data_dir)
])
if args.use_gpu == 1:
exe = fluid.Executor(fluid.CUDAPlace(0))
dataset.set_thread(1)
else:
exe = fluid.Executor(fluid.CPUPlace())
dataset.set_thread(args.num_thread)
exe.run(fluid.default_startup_program())
for epoch_id in range(args.num_epoch):
start = time.time()
sys.stderr.write('\nepoch%d start ...\n' % (epoch_id + 1))
exe.train_from_dataset(
program=main_program,
dataset=dataset,
fetch_list=[loss, auc],
fetch_info=['loss', 'auc'],
debug=False,
print_period=args.print_steps)
model_dir = os.path.join(args.model_output_dir,
'epoch_' + str(epoch_id + 1), "checkpoint")
sys.stderr.write('epoch%d is finished and takes %f s\n' % (
(epoch_id + 1), time.time() - start))
if args.trainer_id == 0: # only trainer 0 save model
print("save model in {}".format(model_dir))
fluid.save(main_program, model_dir)
print("train time cost {:.4f}".format(time.time() - start_time))
print("finish training")
if args.is_local:
print("run local training")
train_loop(fluid.default_main_program())
else:
print("run distribute training")
t = fluid.DistributeTranspiler()
t.transpile(
args.trainer_id, pservers=args.endpoints, trainers=args.trainers)
if args.role == "pserver":
print("run psever")
pserver_prog, pserver_startup = t.get_pserver_programs(
args.current_endpoint)
exe = fluid.Executor(fluid.CPUPlace())
exe.run(pserver_startup)
exe.run(pserver_prog)
elif args.role == "trainer":
print("run trainer")
train_loop(t.get_trainer_program())
if __name__ == "__main__":
utils.check_version()
train()
PaddleRec/ctr/xdeepfm/cluster_train.sh 已删除 100755 → 0
浏览文件 @ 57c0bf10
#!/bin/bash
#export GLOG_v=30
#export GLOG_logtostderr=1
# start pserver0
python -u cluster_train.py \
--train_data_dir data/train_data \
--model_output_dir cluster_model \
--is_local 0 \
--is_sparse \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6000 \
--trainers 2 \
> pserver0.log 2>&1 &
# start pserver1
python -u cluster_train.py \
--train_data_dir data/train_data \
--model_output_dir cluster_model \
--is_local 0 \
--is_sparse \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6001 \
--trainers 2 \
> pserver1.log 2>&1 &
# start trainer0
#CUDA_VISIBLE_DEVICES=1 python cluster_train.py \
python -u cluster_train.py \
--train_data_dir data/train_data \
--model_output_dir cluster_model \
--use_gpu 0 \
--is_local 0 \
--is_sparse \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 0 \
> trainer0.log 2>&1 &
# start trainer1
#CUDA_VISIBLE_DEVICES=2 python cluster_train.py \
python -u cluster_train.py \
--train_data_dir data/train_data \
--model_output_dir cluster_model \
--use_gpu 0 \
--is_local 0 \
--is_sparse \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 1 \
> trainer1.log 2>&1 &
echo "2 pservers and 2 trainers started."
\ No newline at end of file
PaddleRec/ctr/xdeepfm/infer.py
浏览文件 @ a8a58316
......@@ -30,7 +30,7 @@ def infer():
for x in os.listdir(args.test_data_dir)
]
criteo_dataset = CriteoDataset()
test_reader = paddle.batch(
test_reader = fluid.io.batch(
criteo_dataset.test(test_files), batch_size=args.batch_size)
startup_program = fluid.framework.Program()
......
PaddleRec/gru4rec/README.md
浏览文件 @ a8a58316
......@@ -11,8 +11,6 @@
├── infer_sample_neg.py # 预测脚本 sample负例
├── net.py # 网络结构
├── text2paddle.py # 文本数据转paddle数据
├── cluster_train.py # 多机训练
├── cluster_train.sh # 多机训练脚本
├── utils # 通用函数
├── convert_format.py # 转换数据格式
├── vocab.txt # 小样本字典
......@@ -168,7 +166,7 @@ CUDA_VISIBLE_DEVICES=0 python train_sample_neg.py --loss ce --use_cuda 1
可在[net.py](./net.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(
emb = fluid.embedding(
input=src,
size=[vocab_size, hid_size],
param_attr=fluid.ParamAttr(
......@@ -278,12 +276,3 @@ model:model_r@20/epoch_10 recall@20:0.681 time_cost(s):12.2
## 多机训练
厂内用户可以参考[wiki](http://wiki.baidu.com/pages/viewpage.action?pageId=628300529)利用paddlecloud 配置多机环境
可参考cluster_train.py 配置其他多机环境
运行命令本地模拟多机场景, 暂不支持windows
```
sh cluster_train.sh
```
注意本地模拟需要关闭代理
PaddleRec/gru4rec/cluster_train.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
import os
import sys
import time
import six
import numpy as np
import math
import argparse
import paddle.fluid as fluid
import paddle
import time
import utils
import net
SEED = 102
def parse_args():
parser = argparse.ArgumentParser("gru4rec benchmark.")
parser.add_argument(
'--train_dir',
type=str,
default='train_data',
help='train file address')
parser.add_argument(
'--vocab_path',
type=str,
default='vocab.txt',
help='vocab file address')
parser.add_argument('--is_local', type=int, default=1, help='whether local')
parser.add_argument('--hid_size', type=int, default=100, help='hid size')
parser.add_argument(
'--model_dir', type=str, default='model_recall20', help='model dir')
parser.add_argument(
'--batch_size', type=int, default=5, help='num of batch size')
parser.add_argument('--pass_num', type=int, default=10, help='num of epoch')
parser.add_argument(
'--print_batch', type=int, default=10, help='num of print batch')
parser.add_argument(
'--use_cuda', type=int, default=0, help='whether use gpu')
parser.add_argument(
'--base_lr', type=float, default=0.01, help='learning rate')
parser.add_argument(
'--num_devices', type=int, default=1, help='Number of GPU devices')
parser.add_argument(
'--role', type=str, default='pserver', help='trainer or pserver')
parser.add_argument(
'--endpoints',
type=str,
default='127.0.0.1:6000',
help='The pserver endpoints, like: 127.0.0.1:6000, 127.0.0.1:6001')
parser.add_argument(
'--current_endpoint',
type=str,
default='127.0.0.1:6000',
help='The current_endpoint')
parser.add_argument(
'--trainer_id',
type=int,
default=0,
help='trainer id ,only trainer_id=0 save model')
parser.add_argument(
'--trainers',
type=int,
default=1,
help='The num of trianers, (default: 1)')
args = parser.parse_args()
return args
def get_cards(args):
return args.num_devices
def train():
""" do training """
args = parse_args()
hid_size = args.hid_size
train_dir = args.train_dir
vocab_path = args.vocab_path
use_cuda = True if args.use_cuda else False
print("use_cuda:", use_cuda)
batch_size = args.batch_size
vocab_size, train_reader = utils.prepare_data(
train_dir, vocab_path, batch_size=batch_size * get_cards(args),\
buffer_size=1000, word_freq_threshold=0, is_train=True)
# Train program
src_wordseq, dst_wordseq, avg_cost, acc = net.all_vocab_network(
vocab_size=vocab_size, hid_size=hid_size)
# Optimization to minimize lost
sgd_optimizer = fluid.optimizer.SGD(learning_rate=args.base_lr)
sgd_optimizer.minimize(avg_cost)
def train_loop(main_program):
""" train network """
pass_num = args.pass_num
model_dir = args.model_dir
fetch_list = [avg_cost.name]
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
exe = fluid.Executor(place)
exe.run(fluid.default_startup_program())
total_time = 0.0
for pass_idx in six.moves.xrange(pass_num):
epoch_idx = pass_idx + 1
print("epoch_%d start" % epoch_idx)
t0 = time.time()
i = 0
newest_ppl = 0
for data in train_reader():
i += 1
lod_src_wordseq = utils.to_lodtensor([dat[0] for dat in data],
place)
lod_dst_wordseq = utils.to_lodtensor([dat[1] for dat in data],
place)
ret_avg_cost = exe.run(main_program,
feed={
"src_wordseq": lod_src_wordseq,
"dst_wordseq": lod_dst_wordseq
},
fetch_list=fetch_list)
avg_ppl = np.exp(ret_avg_cost[0])
newest_ppl = np.mean(avg_ppl)
if i % args.print_batch == 0:
print("step:%d ppl:%.3f" % (i, newest_ppl))
t1 = time.time()
total_time += t1 - t0
print("epoch:%d num_steps:%d time_cost(s):%f" %
(epoch_idx, i, total_time / epoch_idx))
save_dir = "%s/epoch_%d" % (model_dir, epoch_idx)
feed_var_names = ["src_wordseq", "dst_wordseq"]
fetch_vars = [avg_cost, acc]
if args.trainer_id == 0:
fluid.io.save_inference_model(save_dir, feed_var_names,
fetch_vars, exe)
print("model saved in %s" % save_dir)
print("finish training")
if args.is_local:
print("run local training")
train_loop(fluid.default_main_program())
else:
print("run distribute training")
t = fluid.DistributeTranspiler()
t.transpile(
args.trainer_id, pservers=args.endpoints, trainers=args.trainers)
if args.role == "pserver":
print("run psever")
pserver_prog = t.get_pserver_program(args.current_endpoint)
pserver_startup = t.get_startup_program(args.current_endpoint,
pserver_prog)
exe = fluid.Executor(fluid.CPUPlace())
exe.run(pserver_startup)
exe.run(pserver_prog)
elif args.role == "trainer":
print("run trainer")
train_loop(t.get_trainer_program())
if __name__ == "__main__":
train()
PaddleRec/gru4rec/cluster_train.sh 已删除 100644 → 0
浏览文件 @ 57c0bf10
#!/bin/bash
#export GLOG_v=30
#export GLOG_logtostderr=1
# start pserver0
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--vocab_path vocab.txt \
--batch_size 5 \
--is_local 0 \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6000 \
--trainers 2 \
> pserver0.log 2>&1 &
# start pserver1
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--vocab_path vocab.txt \
--batch_size 5 \
--is_local 0 \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6001 \
--trainers 2 \
> pserver1.log 2>&1 &
# start trainer0
#CUDA_VISIBLE_DEVICES=1 python cluster_train.py \
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--vocab_path vocab.txt \
--batch_size 5 \
--print_batch 10 \
--use_cuda 0 \
--is_local 0 \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 0 \
> trainer0.log 2>&1 &
# start trainer1
#CUDA_VISIBLE_DEVICES=2 python cluster_train.py \
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--vocab_path vocab.txt \
--batch_size 5 \
--print_batch 10 \
--use_cuda 0 \
--is_local 0 \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 1 \
> trainer1.log 2>&1 &
PaddleRec/gru4rec/utils.py
浏览文件 @ a8a58316
......@@ -103,7 +103,7 @@ def prepare_data(file_dir,
if is_train and 'ce_mode' not in os.environ:
vocab_size = get_vocab_size(vocab_path)
reader = sort_batch(
paddle.reader.shuffle(
fluid.io.shuffle(
train(
file_dir, buffer_size, data_type=DataType.SEQ),
buf_size=buffer_size),
......
PaddleRec/multiview_simnet/infer.py
浏览文件 @ a8a58316
......@@ -102,8 +102,8 @@ def parse_args():
def start_infer(args, model_path):
dataset = reader.SyntheticDataset(args.sparse_feature_dim, args.query_slots,
args.title_slots)
test_reader = paddle.batch(
paddle.reader.shuffle(
test_reader = fluid.io.batch(
fluid.io.shuffle(
dataset.valid(), buf_size=args.batch_size * 100),
batch_size=args.batch_size)
place = fluid.CPUPlace()
......
PaddleRec/multiview_simnet/train.py
浏览文件 @ a8a58316
......@@ -112,8 +112,8 @@ def start_train(args):
dataset = reader.SyntheticDataset(args.sparse_feature_dim, args.query_slots,
args.title_slots)
train_reader = paddle.batch(
paddle.reader.shuffle(
train_reader = fluid.io.batch(
fluid.io.shuffle(
dataset.train(), buf_size=args.batch_size * 100),
batch_size=args.batch_size)
place = fluid.CPUPlace()
......
PaddleRec/ncf/evaluate.py
浏览文件 @ a8a58316
import math
import heapq # for retrieval topK
import heapq # for retrieval topK
import multiprocessing
import numpy as np
from time import time
......@@ -23,30 +23,36 @@ _K = None
_args = None
_model_path = None
def run_infer(args, model_path, test_data_path):
test_data_generator = utils.CriteoDataset()
with fluid.scope_guard(fluid.Scope()):
test_reader = paddle.batch(test_data_generator.test(test_data_path, False), batch_size=args.test_batch_size)
test_reader = fluid.io.batch(
test_data_generator.test(test_data_path, False),
batch_size=args.test_batch_size)
place = fluid.CUDAPlace(0) if args.use_gpu else fluid.CPUPlace()
exe = fluid.Executor(place)
exe.run(fluid.default_startup_program())
infer_program, feed_target_names, fetch_vars = fluid.io.load_inference_model(model_path, exe)
infer_program, feed_target_names, fetch_vars = fluid.io.load_inference_model(
model_path, exe)
for data in test_reader():
user_input = np.array([dat[0] for dat in data])
item_input = np.array([dat[1] for dat in data])
pred_val = exe.run(infer_program,
feed={"user_input": user_input,
"item_input": item_input},
fetch_list=fetch_vars,
return_numpy=True)
pred_val = exe.run(
infer_program,
feed={"user_input": user_input,
"item_input": item_input},
fetch_list=fetch_vars,
return_numpy=True)
return pred_val[0].reshape(1, -1).tolist()[0]
def evaluate_model(args, testRatings, testNegatives, K, model_path):
"""
Evaluate the performance (Hit_Ratio, NDCG) of top-K recommendation
......@@ -56,22 +62,23 @@ def evaluate_model(args, testRatings, testNegatives, K, model_path):
global _testRatings
global _testNegatives
global _K
global _model_path
global _model_path
global _args
_args = args
_model_path= model_path
_model_path = model_path
_testRatings = testRatings
_testNegatives = testNegatives
_K = K
hits, ndcgs = [],[]
hits, ndcgs = [], []
for idx in range(len(_testRatings)):
(hr,ndcg) = eval_one_rating(idx)
(hr, ndcg) = eval_one_rating(idx)
hits.append(hr)
ndcgs.append(ndcg)
ndcgs.append(ndcg)
return (hits, ndcgs)
def eval_one_rating(idx):
rating = _testRatings[idx]
items = _testNegatives[idx]
......@@ -80,9 +87,9 @@ def eval_one_rating(idx):
items.append(gtItem)
# Get prediction scores
map_item_score = {}
users = np.full(len(items), u, dtype = 'int32')
users = users.reshape(-1,1)
items_array = np.array(items).reshape(-1,1)
users = np.full(len(items), u, dtype='int32')
users = users.reshape(-1, 1)
items_array = np.array(items).reshape(-1, 1)
temp = np.hstack((users, items_array))
np.savetxt("Data/test.txt", temp, fmt='%d', delimiter=',')
predictions = run_infer(_args, _model_path, _args.test_data_path)
......@@ -91,7 +98,7 @@ def eval_one_rating(idx):
item = items[i]
map_item_score[item] = predictions[i]
items.pop()
# Evaluate top rank list
ranklist = heapq.nlargest(_K, map_item_score, key=map_item_score.get)
hr = getHitRatio(ranklist, gtItem)
......@@ -99,15 +106,17 @@ def eval_one_rating(idx):
return (hr, ndcg)
def getHitRatio(ranklist, gtItem):
for item in ranklist:
if item == gtItem:
return 1
return 0
def getNDCG(ranklist, gtItem):
for i in range(len(ranklist)):
item = ranklist[i]
if item == gtItem:
return math.log(2) / math.log(i+2)
return math.log(2) / math.log(i + 2)
return 0
PaddleRec/ssr/README.md
浏览文件 @ a8a58316
......@@ -43,9 +43,6 @@ cpu 单机多卡训练
CPU_NUM=10 python train.py --train_dir train_data --use_cuda 0 --parallel 1 --batch_size 50 --model_dir model_output --num_devices 10
```
本地模拟多机训练, 不支持windows.
``` bash
sh cluster_train.sh
```
## Inference
......
PaddleRec/ssr/cluster_train.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
#Copyright (c) 2016 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 sys
import time
import argparse
import logging
import paddle.fluid as fluid
import paddle
import utils
import numpy as np
from nets import SequenceSemanticRetrieval
logging.basicConfig(format="%(asctime)s - %(levelname)s - %(message)s")
logger = logging.getLogger("fluid")
logger.setLevel(logging.INFO)
def parse_args():
parser = argparse.ArgumentParser("sequence semantic retrieval")
parser.add_argument(
"--train_dir", type=str, default='train_data', help="Training file")
parser.add_argument(
"--base_lr", type=float, default=0.01, help="learning rate")
parser.add_argument(
'--vocab_path', type=str, default='vocab.txt', help='vocab file')
parser.add_argument(
"--epochs", type=int, default=10, help="Number of epochs")
parser.add_argument(
'--parallel', type=int, default=0, help='whether parallel')
parser.add_argument(
'--use_cuda', type=int, default=0, help='whether use gpu')
parser.add_argument(
'--print_batch', type=int, default=10, help='num of print batch')
parser.add_argument(
'--model_dir', type=str, default='model_output', help='model dir')
parser.add_argument(
"--hidden_size", type=int, default=128, help="hidden size")
parser.add_argument(
"--batch_size", type=int, default=50, help="number of batch")
parser.add_argument(
"--embedding_dim", type=int, default=128, help="embedding dim")
parser.add_argument(
'--num_devices', type=int, default=1, help='Number of GPU devices')
parser.add_argument(
'--step_num', type=int, default=1000, help='Number of steps')
parser.add_argument(
'--enable_ce',
action='store_true',
help='If set, run the task with continuous evaluation logs.')
parser.add_argument(
'--role', type=str, default='pserver', help='trainer or pserver')
parser.add_argument(
'--endpoints',
type=str,
default='127.0.0.1:6000',
help='The pserver endpoints, like: 127.0.0.1:6000, 127.0.0.1:6001')
parser.add_argument(
'--current_endpoint',
type=str,
default='127.0.0.1:6000',
help='The current_endpoint')
parser.add_argument(
'--trainer_id',
type=int,
default=0,
help='trainer id ,only trainer_id=0 save model')
parser.add_argument(
'--trainers',
type=int,
default=1,
help='The num of trianers, (default: 1)')
return parser.parse_args()
def get_cards(args):
return args.num_devices
def train_loop(main_program, avg_cost, acc, train_input_data, place, args,
train_reader):
data_list = [var.name for var in train_input_data]
exe = fluid.Executor(place)
exe.run(fluid.default_startup_program())
loader = fluid.io.DataLoader.from_generator(
feed_list=train_input_data, capacity=10000, iterable=True)
loader.set_sample_list_generator(train_reader, places=place)
train_exe = exe
total_time = 0.0
ce_info = []
for pass_id in range(args.epochs):
epoch_idx = pass_id + 1
print("epoch_%d start" % epoch_idx)
t0 = time.time()
i = 0
for batch_id, data in enumerate(loader()):
i += 1
loss_val, correct_val = train_exe.run(
feed=data, fetch_list=[avg_cost.name, acc.name])
ce_info.append(float(np.mean(correct_val)) / args.batch_size)
if i % args.print_batch == 0:
logger.info(
"Train --> pass: {} batch_id: {} avg_cost: {}, acc: {}".
format(pass_id, batch_id,
np.mean(loss_val),
float(np.mean(correct_val)) / args.batch_size))
if args.enable_ce and i > args.step_num:
break
t1 = time.time()
total_time += t1 - t0
print("epoch:%d num_steps:%d time_cost(s):%f" %
(epoch_idx, i, total_time / epoch_idx))
save_dir = "%s/epoch_%d" % (args.model_dir, epoch_idx)
fluid.save(fluid.default_main_program(), save_dir)
print("model saved in %s" % save_dir)
# only for ce
if args.enable_ce:
ce_acc = 0
try:
ce_acc = ce_info[-2]
except:
print("ce info error")
epoch_idx = args.epochs
device = get_device(args)
if args.use_cuda:
gpu_num = device[1]
print("kpis\teach_pass_duration_gpu%s\t%s" %
(gpu_num, total_time / epoch_idx))
print("kpis\ttrain_acc_gpu%s\t%s" % (gpu_num, ce_acc))
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_acc_cpu%s_thread%s\t%s" %
(cpu_num, threads_num, ce_acc))
def train(args):
if args.enable_ce:
SEED = 102
fluid.default_startup_program().random_seed = SEED
fluid.default_main_program().random_seed = SEED
use_cuda = True if args.use_cuda else False
parallel = True if args.parallel else False
print("use_cuda:", use_cuda, "parallel:", parallel)
train_reader, vocab_size = utils.construct_train_data(
args.train_dir, args.vocab_path, args.batch_size * get_cards(args))
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
ssr = SequenceSemanticRetrieval(vocab_size, args.embedding_dim,
args.hidden_size)
# Train program
train_input_data, cos_pos, avg_cost, acc = ssr.train()
# Optimization to minimize lost
optimizer = fluid.optimizer.Adagrad(learning_rate=args.base_lr)
optimizer.minimize(avg_cost)
print("run distribute training")
t = fluid.DistributeTranspiler()
t.transpile(
args.trainer_id, pservers=args.endpoints, trainers=args.trainers)
if args.role == "pserver":
print("run psever")
pserver_prog = t.get_pserver_program(args.current_endpoint)
pserver_startup = t.get_startup_program(args.current_endpoint,
pserver_prog)
exe = fluid.Executor(fluid.CPUPlace())
exe.run(pserver_startup)
exe.run(pserver_prog)
elif args.role == "trainer":
print("run trainer")
train_loop(t.get_trainer_program(), avg_cost, acc, train_input_data,
place, args, train_reader)
def get_device(args):
if args.use_cuda:
gpus = os.environ.get("CUDA_VISIBLE_DEVICES", 1)
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)
def main():
args = parse_args()
train(args)
if __name__ == "__main__":
main()
PaddleRec/ssr/cluster_train.sh 已删除 100644 → 0
浏览文件 @ 57c0bf10
#!/bin/bash
#export GLOG_v=30
#export GLOG_logtostderr=1
# start pserver0
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--vocab_path vocab.txt \
--batch_size 5 \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6000 \
--trainers 2 \
> pserver0.log 2>&1 &
# start pserver1
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--vocab_path vocab.txt \
--batch_size 5 \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6001 \
--trainers 2 \
> pserver1.log 2>&1 &
# start trainer0
#CUDA_VISIBLE_DEVICES=1 python cluster_train.py \
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--vocab_path vocab.txt \
--batch_size 5 \
--print_batch 10 \
--use_cuda 0 \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 0 \
> trainer0.log 2>&1 &
# start trainer1
#CUDA_VISIBLE_DEVICES=2 python cluster_train.py \
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--vocab_path vocab.txt \
--batch_size 5 \
--print_batch 10 \
--use_cuda 0 \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 1 \
> trainer1.log 2>&1 &
PaddleRec/ssr/infer.py
浏览文件 @ a8a58316
......@@ -37,16 +37,16 @@ def parse_args():
def model(vocab_size, emb_size, hidden_size):
user_data = fluid.layers.data(
name="user", shape=[1], dtype="int64", lod_level=1)
all_item_data = fluid.layers.data(
name="all_item", shape=[vocab_size, 1], dtype="int64")
user_data = fluid.data(
name="user", shape=[None, 1], dtype="int64", lod_level=1)
all_item_data = fluid.data(
name="all_item", shape=[None, vocab_size, 1], dtype="int64")
user_emb = fluid.layers.embedding(
user_emb = fluid.embedding(
input=user_data, size=[vocab_size, emb_size], param_attr="emb.item")
all_item_emb = fluid.layers.embedding(
all_item_emb = fluid.embedding(
input=all_item_data, size=[vocab_size, emb_size], param_attr="emb.item")
all_item_emb_re = fluid.layers.reshape(x=all_item_emb, shape=[-1, emb_size])
all_item_emb_re = all_item_emb
user_encoder = net.GrnnEncoder(hidden_size=hidden_size)
user_enc = user_encoder.forward(user_emb)
......@@ -63,7 +63,7 @@ def model(vocab_size, emb_size, hidden_size):
bias_attr="item.b")
cos_item = fluid.layers.cos_sim(X=all_item_hid, Y=user_re)
all_pre_ = fluid.layers.reshape(x=cos_item, shape=[-1, vocab_size])
pos_label = fluid.layers.data(name="pos_label", shape=[1], dtype="int64")
pos_label = fluid.data(name="pos_label", shape=[None, 1], dtype="int64")
acc = fluid.layers.accuracy(input=all_pre_, label=pos_label, k=20)
return acc
......
PaddleRec/ssr/utils.py
浏览文件 @ a8a58316
......@@ -18,7 +18,7 @@ def construct_train_data(file_dir, vocab_path, batch_size):
files = [file_dir + '/' + f for f in os.listdir(file_dir)]
y_data = reader.YoochooseDataset(vocab_size)
train_reader = fluid.io.batch(
paddle.reader.shuffle(
fluid.io.shuffle(
y_data.train(files), buf_size=batch_size * 100),
batch_size=batch_size)
return train_reader, vocab_size
......
PaddleRec/tagspace/README.md
浏览文件 @ a8a58316
......@@ -9,8 +9,6 @@
├── infer.py # 预测脚本
├── net.py # 网络结构
├── text2paddle.py # 文本数据转paddle数据
├── cluster_train.py # 多机训练
├── cluster_train.sh # 多机训练脚本
├── utils # 通用函数
├── vocab_text.txt # 小样本文本字典
├── vocab_tag.txt # 小样本类别字典
......@@ -89,9 +87,3 @@ python infer.py
```
python infer.py --model_dir big_model --vocab_tag_path big_vocab_tag.txt --test_dir test_big_data/
```
## 本地模拟多机
运行命令
```
sh cluster_train.py
```
PaddleRec/tagspace/cluster_train.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
import os
import sys
import time
import six
import numpy as np
import math
import argparse
import paddle
import paddle.fluid as fluid
import time
import utils
import net
SEED = 102
def parse_args():
parser = argparse.ArgumentParser("TagSpace benchmark.")
parser.add_argument(
'--neg_size', type=int, default=3, help='neg/pos ratio')
parser.add_argument(
'--train_dir', type=str, default='train_data', help='train file address')
parser.add_argument(
'--vocab_text_path', type=str, default='vocab_text.txt', help='vocab_text file address')
parser.add_argument(
'--vocab_tag_path', type=str, default='vocab_tag.txt', help='vocab_text file address')
parser.add_argument(
'--is_local', type=int, default=1, help='whether local')
parser.add_argument(
'--model_dir', type=str, default='model_', help='model dir')
parser.add_argument(
'--batch_size', type=int, default=5, help='num of batch size')
parser.add_argument(
'--print_batch', type=int, default=10, help='num of print batch')
parser.add_argument(
'--pass_num', type=int, default=10, help='num of epoch')
parser.add_argument(
'--use_cuda', type=int, default=0, help='whether use gpu')
parser.add_argument(
'--base_lr', type=float, default=0.01, help='learning rate')
parser.add_argument(
'--num_devices', type=int, default=1, help='Number of GPU devices')
parser.add_argument(
'--role', type=str, default='pserver', help='trainer or pserver')
parser.add_argument(
'--endpoints', type=str, default='127.0.0.1:6000', help='The pserver endpoints, like: 127.0.0.1:6000, 127.0.0.1:6001')
parser.add_argument(
'--current_endpoint', type=str, default='127.0.0.1:6000', help='The current_endpoint')
parser.add_argument(
'--trainer_id', type=int, default=0, help='trainer id ,only trainer_id=0 save model')
parser.add_argument(
'--trainers', type=int, default=1, help='The num of trianers, (default: 1)')
args = parser.parse_args()
return args
def get_cards(args):
return args.num_devices
def train():
""" do training """
args = parse_args()
train_dir = args.train_dir
vocab_text_path = args.vocab_text_path
vocab_tag_path = args.vocab_tag_path
use_cuda = True if args.use_cuda else False
batch_size = args.batch_size
neg_size = args.neg_size
vocab_text_size, vocab_tag_size, train_reader = utils.prepare_data(
file_dir=train_dir, vocab_text_path=vocab_text_path,
vocab_tag_path=vocab_tag_path, neg_size=neg_size,
batch_size=batch_size * get_cards(args),
buffer_size=batch_size*100, is_train=True)
""" train network """
# Train program
avg_cost, correct, cos_pos = net.network(vocab_text_size, vocab_tag_size, neg_size=neg_size)
# Optimization to minimize lost
sgd_optimizer = fluid.optimizer.SGD(learning_rate=args.base_lr)
sgd_optimizer.minimize(avg_cost)
def train_loop(main_program):
# Initialize executor
place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
exe = fluid.Executor(place)
pass_num = args.pass_num
model_dir = args.model_dir
fetch_list = [avg_cost.name]
exe.run(fluid.default_startup_program())
total_time = 0.0
for pass_idx in range(pass_num):
epoch_idx = pass_idx + 1
print("epoch_%d start" % epoch_idx)
t0 = time.time()
for batch_id, data in enumerate(train_reader()):
lod_text_seq = utils.to_lodtensor([dat[0] for dat in data], place)
lod_pos_tag = utils.to_lodtensor([dat[1] for dat in data], place)
lod_neg_tag = utils.to_lodtensor([dat[2] for dat in data], place)
loss_val, correct_val = exe.run(
feed={
"text": lod_text_seq,
"pos_tag": lod_pos_tag,
"neg_tag": lod_neg_tag},
fetch_list=[avg_cost.name, correct.name])
if batch_id % args.print_batch == 0:
print("TRAIN --> pass: {} batch_num: {} avg_cost: {}, acc: {}"
.format(pass_idx, (batch_id+10) * batch_size, np.mean(loss_val),
float(np.sum(correct_val)) / batch_size))
t1 = time.time()
total_time += t1 - t0
print("epoch:%d num_steps:%d time_cost(s):%f" %
(epoch_idx, batch_id, total_time / epoch_idx))
save_dir = "%s/epoch_%d" % (model_dir, epoch_idx)
feed_var_names = ["text", "pos_tag"]
fetch_vars = [cos_pos]
fluid.io.save_inference_model(save_dir, feed_var_names, fetch_vars, exe)
print("finish training")
if args.is_local:
print("run local training")
train_loop(fluid.default_main_program())
else:
print("run distribute training")
t = fluid.DistributeTranspiler()
t.transpile(args.trainer_id, pservers=args.endpoints, trainers=args.trainers)
if args.role == "pserver":
print("run psever")
pserver_prog = t.get_pserver_program(args.current_endpoint)
pserver_startup = t.get_startup_program(args.current_endpoint,
pserver_prog)
exe = fluid.Executor(fluid.CPUPlace())
exe.run(pserver_startup)
exe.run(pserver_prog)
elif args.role == "trainer":
print("run trainer")
train_loop(t.get_trainer_program())
if __name__ == "__main__":
train()
PaddleRec/tagspace/cluster_train.sh 已删除 100644 → 0
浏览文件 @ 57c0bf10
#!/bin/bash
#export GLOG_v=30
#export GLOG_logtostderr=1
# start pserver0
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--batch_size 5 \
--is_local 0 \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6000 \
--trainers 2 \
> pserver0.log 2>&1 &
# start pserver1
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--batch_size 5 \
--is_local 0 \
--role pserver \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--current_endpoint 127.0.0.1:6001 \
--trainers 2 \
> pserver1.log 2>&1 &
# start trainer0
#CUDA_VISIBLE_DEVICES=1 python cluster_train.py \
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--batch_size 5 \
--print_batch 10 \
--use_cuda 0 \
--is_local 0 \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 0 \
> trainer0.log 2>&1 &
# start trainer1
#CUDA_VISIBLE_DEVICES=2 python cluster_train.py \
python cluster_train.py \
--train_dir train_data \
--model_dir cluster_model \
--batch_size 5 \
--print_batch 10 \
--use_cuda 0 \
--is_local 0 \
--role trainer \
--endpoints 127.0.0.1:6000,127.0.0.1:6001 \
--trainers 2 \
--trainer_id 1 \
> trainer1.log 2>&1 &
PaddleRec/tagspace/utils.py
浏览文件 @ a8a58316
......@@ -65,7 +65,7 @@ def prepare_data(file_dir,
vocab_text_size = get_vocab_size(vocab_text_path)
vocab_tag_size = get_vocab_size(vocab_tag_path)
reader = sort_batch(
paddle.reader.shuffle(
fluid.io.shuffle(
train(
file_dir,
vocab_tag_size,
......
PaddleRec/tdm/tdm_demo/README.md
浏览文件 @ a8a58316
......@@ -51,7 +51,7 @@
- **Node-Embedding**：注意，此处的Embedding，并非我们已有的item-embedding，而是构建完成的树的节点对应的Embedding，由item-embedding通过规则生成，是我们的网络主要训练的目标。ID范围为所有0->节点数-1。我们同时也需准备一个映射表，来告诉模型，item_id到node_id的映射关系。
- **Travel**：是指叶子节点从root开始直到其自身的遍历路径，如上图，14号节点的Travel：0->1->3->7->14
- **Layer**：指树的层，如上图，共有4层。
> Paddle-TDM在训练时，不会改动树的结构，只会改动Node-Embedding。
......@@ -156,7 +156,7 @@ TDM的组网，宏观上，可以概括为三个部分
**demo模型，假设输入为两个元素：**
> 一、user/query的emb表示，该emb应该来源于特征的组合在某个空间的映射（比如若干特征取emb后concat到一起），或其他预训练模型的处理结果（比如将明文query通过nlp预处理得到emb表示）
> 二、item的正样本，是发生了实际点击/购买/浏览等行为的item_id，与输入的user/query emb强相关，是我们之后通过预测想得到的结果。
在paddle组网中，我们这样定义上面两个变量：
......@@ -233,9 +233,9 @@ tdm_sampler的运行逻辑如下：
- 在item遍历路径上的node视为正样本，`positive_node_id`由`travel_list[item_id][i]`给出，其他同层的兄弟节点视为负样本，该层节点列表由`layer_list[i]`给出，如果`positive_node_id`不在`layer_list[i]`中，会提示错误。
- 在兄弟节点中进行随机采样，采样N个node，N由`neg_sampling_list[i]`的值决定，如果该值大于兄弟节点的数量，会提示错误。 采样结果不会重复，且不会采样到正样本。
- 如果`output_positive=True`，则会同时输出正负样本，否则只输出负采样的结果
- 生成该层`label`，shape与采样结果一致，正样本对应的label=1，负样本的label=0
- 生成该层`mask`，如果树是不平衡的，则有些item不会位于树的最后一层，所以遍历路径的实际长度会比其他item少，为了tensor维度一致，travel_list中padding了0。当遇到了padding的0时，tdm_sampler也会输出正常维度的采样结果，采样结果与label都为0。为了区分这部分虚拟的采样结果与真实采样结果，会给虚拟采样结果额外设置mask=0，如果是真实采样结果mask=1
......@@ -403,23 +403,23 @@ acc = fluid.layers.accuracy(input=softmax_prob, label=labels_reshape)
在demo网络中，我们设置为从某一层的所有节点开始进行检索。paddle组网对输入定义的实现如下：
```python
def input_data(self):
input_emb = fluid.layers.data(
input_emb = fluid.data(
name="input_emb",
shape=[self.input_embed_size],
shape=[None, self.input_embed_size],
dtype="float32",
)
# first_layer 与 first_layer_mask 对应着infer起始的节点
first_layer = fluid.layers.data(
first_layer = fluid.data(
name="first_layer_node",
shape=[1],
shape=[None, 1],
dtype="int64",
lod_level=1, #支持变长
)
first_layer_mask = fluid.layers.data(
first_layer_mask = fluid.data(
name="first_layer_node_mask",
shape=[1],
shape=[None, 1],
dtype="int64",
lod_level=1,
)
......@@ -447,7 +447,7 @@ def create_first_layer(self, args):
tdm的检索逻辑类似beamsearch，简单来说：在每一层计算打分，得到topK的节点，将这些节点的孩子节点作为下一层的输入，如此循环，得到最终的topK。但仍然有一些需要注意的细节，下面将详细介绍。
- 问题一：怎么处理`input_emb`？
- input_emb过`input_fc`，检索中，只需过一次即可:
```python
nput_trans_emb = self.input_trans_net.input_fc_infer(input_emb)
......@@ -663,7 +663,7 @@ if args.save_init_model or not args.load_model:
```
> 为什么每次加载模型手动Set `learning rate`？
>
>
> 学习率在paddle的组网中，是以一个`persistable=Ture`的长期变量储存在模型的Variable scope里的。每次使用load_persistables加载模型时，也会使用加载的模型的学习率覆盖本地模型的默认学习率，换言之，加载init_model以后，学习率也是保存init_model时的学习率。对模型的调试会产生不必要的影响，为了保证网络训练如预期，需要这样的手动set步骤。
### demo的训练运行方法
......@@ -706,4 +706,4 @@ Demo代码中给出了基于paddle预测库加载tdm模型，输入emb产出item
```
- 首先需要运行`run_infer.sh`，打开`save_init_model`开关，使用`save_inference_model`产出paddle的推理模型，`predict.py`会加载`infer_model`，进行高速推理。
- 欲想进一步高速推理，需使用含预测库的paddle预测库，可以使用`mkl`及`mkl_dnn`等计算库加速op的计算。相关文档可以参考：[服务器端部署](https://www.paddlepaddle.org.cn/documentation/docs/zh/advanced_guide/inference_deployment/inference/index_cn.html)
- tdm相关op目前仅支持在cpu设备上运行，后续会支持GPU，欢迎关注。
\ No newline at end of file
- tdm相关op目前仅支持在cpu设备上运行，后续会支持GPU，欢迎关注。
PaddleRec/tdm/tdm_demo/dataset_generator.py
浏览文件 @ a8a58316
......@@ -35,6 +35,7 @@ class TDMDataset(dg.MultiSlotStringDataGenerator):
"""
Read test_data line by line & yield batch
"""
def local_iter():
"""Read file line by line"""
for fname in infer_file_list:
......@@ -46,13 +47,14 @@ class TDMDataset(dg.MultiSlotStringDataGenerator):
yield [input_emb]
import paddle
batch_iter = paddle.batch(local_iter, batch)
batch_iter = fluid.io.batch(local_iter, batch)
return batch_iter
def generate_sample(self, line):
"""
Read the data line by line and process it as a dictionary
"""
def iterator():
"""
This function needs to be implemented by the user, based on data format
......
PaddleRec/tdm/tdm_demo/infer_network.py
浏览文件 @ a8a58316
......@@ -41,26 +41,23 @@ class TdmInferNet(object):
self.input_trans_net = InputTransNet(args)
def input_data(self):
input_emb = fluid.layers.data(
input_emb = fluid.data(
name="input_emb",
shape=[self.input_embed_size],
dtype="float32",
)
shape=[None, self.input_embed_size],
dtype="float32", )
# first_layer 与 first_layer_mask 对应着infer起始层的节点
first_layer = fluid.layers.data(
first_layer = fluid.data(
name="first_layer_node",
shape=[1],
shape=[None, 1],
dtype="int64",
lod_level=1,
)
lod_level=1, )
first_layer_mask = fluid.layers.data(
first_layer_mask = fluid.data(
name="first_layer_node_mask",
shape=[1],
shape=[None, 1],
dtype="int64",
lod_level=1,
)
lod_level=1, )
inputs = [input_emb] + [first_layer] + [first_layer_mask]
return inputs
......@@ -125,28 +122,27 @@ class TdmInferNet(object):
size=[self.node_nums, self.node_embed_size],
param_attr=fluid.ParamAttr(name="TDM_Tree_Emb"))
input_fc_out = self.input_trans_net.layer_fc_infer(
input_trans_emb, layer_idx)
input_fc_out = self.input_trans_net.layer_fc_infer(input_trans_emb,
layer_idx)
# 过每一层的分类器
layer_classifier_res = self.layer_classifier.classifier_layer_infer(input_fc_out,
node_emb,
layer_idx)
layer_classifier_res = self.layer_classifier.classifier_layer_infer(
input_fc_out, node_emb, layer_idx)
# 过最终的判别分类器
tdm_fc = fluid.layers.fc(input=layer_classifier_res,
size=self.label_nums,
act=None,
num_flatten_dims=2,
param_attr=fluid.ParamAttr(
name="tdm.cls_fc.weight"),
bias_attr=fluid.ParamAttr(name="tdm.cls_fc.bias"))
tdm_fc = fluid.layers.fc(
input=layer_classifier_res,
size=self.label_nums,
act=None,
num_flatten_dims=2,
param_attr=fluid.ParamAttr(name="tdm.cls_fc.weight"),
bias_attr=fluid.ParamAttr(name="tdm.cls_fc.bias"))
prob = fluid.layers.softmax(tdm_fc)
positive_prob = fluid.layers.slice(
prob, axes=[2], starts=[1], ends=[2])
prob_re = fluid.layers.reshape(
positive_prob, [-1, current_layer_node_num])
prob_re = fluid.layers.reshape(positive_prob,
[-1, current_layer_node_num])
# 过滤掉padding产生的无效节点（node_id=0）
node_zero_mask = fluid.layers.cast(current_layer_node, 'bool')
......@@ -161,11 +157,10 @@ class TdmInferNet(object):
# index_sample op根据下标索引tensor对应位置的值
# 若paddle版本>2.0，调用方式为paddle.index_sample
top_node = fluid.contrib.layers.index_sample(
current_layer_node, topk_i)
top_node = fluid.contrib.layers.index_sample(current_layer_node,
topk_i)
prob_re_mask = prob_re * current_layer_child_mask # 过滤掉非叶子节点
topk_value = fluid.contrib.layers.index_sample(
prob_re_mask, topk_i)
topk_value = fluid.contrib.layers.index_sample(prob_re_mask, topk_i)
node_score.append(topk_value)
node_list.append(top_node)
......@@ -190,7 +185,8 @@ class TdmInferNet(object):
res_node = fluid.layers.reshape(res_layer_node, [-1, self.topK, 1])
# 利用Tree_info信息，将node_id转换为item_id
tree_info = fluid.default_main_program().global_block().var("TDM_Tree_Info")
tree_info = fluid.default_main_program().global_block().var(
"TDM_Tree_Info")
res_node_emb = fluid.layers.gather_nd(tree_info, res_node)
res_item = fluid.layers.slice(
......
PaddleRec/text_matching_on_quora/.run_ce.sh 已删除 100755 → 0
浏览文件 @ 57c0bf10
#!/bin/bash
export MKL_NUM_THREADS=1
export OMP_NUM_THREADS=1
cudaid=${text_matching_on_quora:=0} # use 0-th card as default
export CUDA_VISIBLE_DEVICES=$cudaid
FLAGS_benchmark=true python train_and_evaluate.py --model_name=cdssmNet --config=cdssm_base --enable_ce --epoch_num=5 | python _ce.py
cudaid=${text_matching_on_quora_m:=0,1,2,3} # use 0,1,2,3 card as default
export CUDA_VISIBLE_DEVICES=$cudaid
FLAGS_benchmark=true python train_and_evaluate.py --model_name=cdssmNet --config=cdssm_base --enable_ce --epoch_num=5 | python _ce.py
PaddleRec/text_matching_on_quora/README.md 已删除 100644 → 0
浏览文件 @ 57c0bf10
# Text matching on Quora qestion-answer pair dataset
## contents
* [Introduction](#introduction)
* [a brief review of the Quora Question Pair (QQP) Task](#a-brief-review-of-the-quora-question-pair-qqp-task)
* [Our Work](#our-work)
* [Environment Preparation](#environment-preparation)
* [Install Fluid release 1.0](#install-fluid-release-10)
* [cpu version](#cpu-version)
* [gpu version](#gpu-version)
* [Have I installed Fluid successfully?](#have-i-installed-fluid-successfully)
* [Prepare Data](#prepare-data)
* [Train and evaluate](#train-and-evaluate)
* [Models](#models)
* [Results](#results)
## Introduction
### a brief review of the Quora Question Pair (QQP) Task
The [Quora Question Pair](https://data.quora.com/First-Quora-Dataset-Release-Question-Pairs) dataset contains 400,000 question pairs from [Quora](https://www.quora.com/), where people ask and answer questions related to specific areas. Each sample in the dataset consists of two questions (both English) and a label that represents whether the questions are duplicate. The dataset is well annotated by human.
Below are two samples from the dataset. The last column indicates whether the two questions are duplicate (1) or not (0).
|id | qid1 | qid2| question1| question2| is_duplicate
|:---:|:---:|:---:|:---:|:---:|:---:|
|0 |1 |2 |What is the step by step guide to invest in share market in india? |What is the step by step guide to invest in share market? |0|
|1 |3 |4 |What is the story of Kohinoor (Koh-i-Noor) Diamond? | What would happen if the Indian government stole the Kohinoor (Koh-i-Noor) diamond back? |0|
A [kaggle competition](https://www.kaggle.com/c/quora-question-pairs#description) was held based on this dataset in 2017. The kagglers were given a training dataset (with labels), and requested to make predictions on a test dataset (without labels). The predictions were evaluated by the log-likelihood loss on the test data.
The kaggle competition has inspired much effective work. However, most of these models are rule-based and difficult to be transferred to new tasks. Researchers are seeking for more general models that work well on this task and other natual language processing (NLP) tasks.
[Wang _et al._](https://arxiv.org/abs/1702.03814) proposed a bilateral multi-perspective matching (BIMPM) model based on the Quora Question Pair dataset. They splitted the original dataset to [3 parts](https://drive.google.com/file/d/0B0PlTAo--BnaQWlsZl9FZ3l1c28/view?usp=sharing): _train.tsv_ (384,348 samples), _dev.tsv_ (10,000 samples) and _test.tsv_ (10,000 samples). The class distribution of _train.tsv_ is unbalanced (37% positive and 63% negative), while those of _dev.tsv_ and _test.tsv_ are balanced(50% positive and 50% negetive). We used the same splitting method in our experiments.
### Our Work
Based on the Quora Question Pair Dataset, we implemented some classic models in the area of neural language understanding (NLU). The accuracy of prediction results are evaluated on the _test.tsv_ from [Wang _et al._](https://arxiv.org/abs/1702.03814).
## Environment Preparation
### Install Fluid release 1.0
Please follow the [official document in English](http://www.paddlepaddle.org/documentation/docs/en/1.0/build_and_install/pip_install_en.html) or [official document in Chinese](http://www.paddlepaddle.org/documentation/docs/zh/1.0/beginners_guide/install/Start.html) to install the Fluid deep learning framework.
#### Have I installed Fluid successfully?
Run the following script from your command line:
```shell
python -c "import paddle"
```
If Fluid is installed successfully you should see no error message. Feel free to open issues under the [PaddlePaddle repository](https://github.com/PaddlePaddle/Paddle/issues) for support.
## Prepare Data
Please download the Quora dataset from [Google drive](https://drive.google.com/file/d/0B0PlTAo--BnaQWlsZl9FZ3l1c28/view?usp=sharing) and unzip to $HOME/.cache/paddle/dataset.
Then run _data/prepare_quora_data.sh_ to download the pre-trained _word2vec_ embedding file -- _glove.840B.300d.zip_:
```shell
sh data/prepare_quora_data.sh
```
At this point the dataset directory ($HOME/.cache/paddle/dataset) structure should be:
```shell
$HOME/.cache/paddle/dataset
|- Quora_question_pair_partition
|- train.tsv
|- test.tsv
|- dev.tsv
|- readme.txt
|- wordvec.txt
|- glove.840B.300d.txt
```
## Train and evaluate
We provide multiple models and configurations. Details are shown in `models` and `configs` directories. For a quick start, please run the _cdssmNet_ model with the corresponding configuration:
```shell
python train_and_evaluate.py \
--model_name=cdssmNet \
--config=cdssm_base
```
Logs will be output to the console. If everything works well, the logging information will have the same formats as the content in _cdssm_base.log_.
All configurations used in our experiments are as follows:
|Model|Config|command
|:----:|:----:|:----:|
|cdssmNet|cdssm_base|python train_and_evaluate.py --model_name=cdssmNet --config=cdssm_base
|DecAttNet|decatt_glove|python train_and_evaluate.py --model_name=DecAttNet --config=decatt_glove
|InferSentNet|infer_sent_v1|python train_and_evaluate.py --model_name=InferSentNet --config=infer_sent_v1
|InferSentNet|infer_sent_v2|python train_and_evaluate.py --model_name=InferSentNet --config=infer_sent_v2
|SSENet|sse_base|python train_and_evaluate.py --model_name=SSENet --config=sse_base
## Models
We implemeted 4 models for now: the convolutional deep-structured semantic model (CDSSM, CNN-based), the InferSent model (RNN-based), the shortcut-stacked encoder (SSE, RNN-based), and the decomposed attention model (DecAtt, attention-based).
|Model|features|Context Encoder|Match Layer|Classification Layer
|:----:|:----:|:----:|:----:|:----:|
|CDSSM|word|1 layer conv1d|concatenation|MLP
|DecAtt|word|Attention|concatenation|MLP
|InferSent|word|1 layer Bi-LSTM|concatenation/element-wise product/<br>absolute element-wise difference|MLP
|SSE|word|3 layer Bi-LSTM|concatenation/element-wise product/<br>absolute element-wise difference|MLP
### CDSSM
```
@inproceedings{shen2014learning,
title={Learning semantic representations using convolutional neural networks for web search},
author={Shen, Yelong and He, Xiaodong and Gao, Jianfeng and Deng, Li and Mesnil, Gr{\'e}goire},
booktitle={Proceedings of the 23rd International Conference on World Wide Web},
pages={373--374},
year={2014},
organization={ACM}
}
```
### InferSent
```
@article{conneau2017supervised,
title={Supervised learning of universal sentence representations from natural language inference data},
author={Conneau, Alexis and Kiela, Douwe and Schwenk, Holger and Barrault, Loic and Bordes, Antoine},
journal={arXiv preprint arXiv:1705.02364},
year={2017}
}
```
### SSE
```
@article{nie2017shortcut,
title={Shortcut-stacked sentence encoders for multi-domain inference},
author={Nie, Yixin and Bansal, Mohit},
journal={arXiv preprint arXiv:1708.02312},
year={2017}
}
```
### DecAtt
```
@article{tomar2017neural,
title={Neural paraphrase identification of questions with noisy pretraining},
author={Tomar, Gaurav Singh and Duque, Thyago and T{\"a}ckstr{\"o}m, Oscar and Uszkoreit, Jakob and Das, Dipanjan},
journal={arXiv preprint arXiv:1704.04565},
year={2017}
}
```
## Results
|Model|Config|dev accuracy| test accuracy
|:----:|:----:|:----:|:----:|
|cdssmNet|cdssm_base|83.56%|82.83%|
|DecAttNet|decatt_glove|86.31%|86.22%|
|InferSentNet|infer_sent_v1|87.15%|86.62%|
|InferSentNet|infer_sent_v2|88.55%|88.43%|
|SSENet|sse_base|88.35%|88.25%|
In our experiment, we found that LSTM-based models outperformed convolution-based models. The DecAtt model has fewer parameters than LSTM-based models, but is sensitive to hyper-parameters.
<p align="center">
<img src="imgs/models_test_acc.png" width = "500" alt="test_acc"/>
</p>
PaddleRec/text_matching_on_quora/_ce.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
# 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
each_pass_duration_card1_kpi = DurationKpi(
'each_pass_duration_card1', 0.08, 0, actived=True)
train_avg_cost_card1_kpi = CostKpi('train_avg_cost_card1', 0.08, 0)
train_avg_acc_card1_kpi = CostKpi('train_avg_acc_card1', 0.02, 0)
each_pass_duration_card4_kpi = DurationKpi(
'each_pass_duration_card4', 0.08, 0, actived=True)
train_avg_cost_card4_kpi = CostKpi('train_avg_cost_card4', 0.08, 0)
train_avg_acc_card4_kpi = CostKpi('train_avg_acc_card4', 0.02, 0)
tracking_kpis = [
each_pass_duration_card1_kpi,
train_avg_cost_card1_kpi,
train_avg_acc_card1_kpi,
each_pass_duration_card4_kpi,
train_avg_cost_card4_kpi,
train_avg_acc_card4_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)
PaddleRec/text_matching_on_quora/configs/__init__.py 已删除 100755 → 0
浏览文件 @ 57c0bf10
# Copyright (c) 2016 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.
from .cdssm import cdssm_base
from .dec_att import decatt_glove
from .sse import sse_base
from .infer_sent import infer_sent_v1
from .infer_sent import infer_sent_v2
PaddleRec/text_matching_on_quora/imgs/README.md 已删除 100644 → 0
浏览文件 @ 57c0bf10
Image files for this model: text_matching_on_quora
PaddleRec/text_matching_on_quora/metric.py 已删除 100755 → 0
浏览文件 @ 57c0bf10
此差异已折叠。
PaddleRec/text_matching_on_quora/utils.py 已删除 100755 → 0
浏览文件 @ 57c0bf10
此差异已折叠。
PaddleRec/word2vec/README.md
浏览文件 @ a8a58316
此差异已折叠。
PaddleRec/word2vec/cluster_train.py 已删除 100644 → 0
浏览文件 @ 57c0bf10
此差异已折叠。
PaddleRec/word2vec/cluster_train.sh 已删除 100644 → 0
浏览文件 @ 57c0bf10
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