Huawei_Technology
/
mindspore
Not watched
1
0
Fork 0
Code
Releases 13 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
Browse Source

add DeepFM

tags/v0.3.0-alpha
yangyongjie 5 years ago
parent
a8efea5c81
commit
8f79f0cce8
12 changed files with 1265 additions and 0 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +132
    -0
      example/deepfm_criteo/README.md
  2. +14
    -0
      example/deepfm_criteo/__init__.py
  3. +66
    -0
      example/deepfm_criteo/eval.py
  4. +44
    -0
      example/deepfm_criteo/scripts/run_distribute_train.sh
  5. +32
    -0
      example/deepfm_criteo/scripts/run_eval.sh
  6. +34
    -0
      example/deepfm_criteo/scripts/run_standalone_train.sh
  7. +14
    -0
      example/deepfm_criteo/src/__init__.py
  8. +107
    -0
      example/deepfm_criteo/src/callback.py
  9. +62
    -0
      example/deepfm_criteo/src/config.py
  10. +299
    -0
      example/deepfm_criteo/src/dataset.py
  11. +370
    -0
      example/deepfm_criteo/src/deepfm.py
  12. +91
    -0
      example/deepfm_criteo/train.py

+ 132
- 0
example/deepfm_criteo/README.md View File

@@ -0,0 +1,132 @@
# DeepFM Description

This is an example of training DeepFM with Criteo dataset in MindSpore.

[Paper](https://arxiv.org/pdf/1703.04247.pdf) Huifeng Guo, Ruiming Tang, Yunming Ye, Zhenguo Li, Xiuqiang He


# Model architecture

The overall network architecture of DeepFM is show below:

[Link](https://arxiv.org/pdf/1703.04247.pdf)


# Requirements
- Install [MindSpore](https://www.mindspore.cn/install/en).
- Download the criteo dataset for pre-training. Extract and clean text in the dataset with [WikiExtractor](https://github.com/attardi/wikiextractor). Convert the dataset to TFRecord format and move the files to a specified path.
- For more information, please check the resources below：
- [MindSpore tutorials](https://www.mindspore.cn/tutorial/zh-CN/master/index.html)
- [MindSpore API](https://www.mindspore.cn/api/zh-CN/master/index.html)

# Script description

## Script and sample code

```python
├── deepfm
├── README.md
├── scripts
│ ├──run_train.sh
│ ├──run_eval.sh
├── src
│ ├──config.py
│ ├──dataset.py
│ ├──callback.py
│ ├──deepfm.py
├── train.py
├── eval.py
```

## Training process

### Usage

- sh run_train.sh [DEVICE_NUM] [DATASET_PATH] [MINDSPORE_HCCL_CONFIG_PAHT]
- python train.py --dataset_path [DATASET_PATH]

### Launch

```
# distribute training example
sh scripts/run_distribute_train.sh 8 /opt/dataset/criteo /opt/mindspore_hccl_file.json
# standalone training example
sh scripts/run_standalone_train.sh 0 /opt/dataset/criteo
or
python train.py --dataset_path /opt/dataset/criteo > output.log 2>&1 &
```

### Result

Training result will be stored in the example path.
Checkpoints will be stored at `./checkpoint` by default,
and training log will be redirected to `./output.log` by default,
and loss log will be redirected to `./loss.log` by default,
and eval log will be redirected to `./auc.log` by default.


## Eval process

### Usage

- sh run_eval.sh [DEVICE_ID] [DATASET_PATH] [CHECKPOINT_PATH]

### Launch

```
# infer example
sh scripts/run_eval.sh 0 ~/criteo/eval/ ~/train/deepfm-15_41257.ckpt
```

> checkpoint can be produced in training process.

### Result

Inference result will be stored in the example path, you can find result like the followings in `auc.log`.

```
2020-05-27 20:51:35 AUC: 0.80577889065281, eval time: 35.55999s.
```

# Model description

## Performance

### Training Performance

| Parameters | DeepFM |
| -------------------------- | ------------------------------------------------------|
| Model Version | |
| Resource | Ascend 910, cpu:2.60GHz 96cores, memory:1.5T |
| uploaded Date | 05/27/2020 |
| MindSpore Version | 0.2.0 |
| Dataset | Criteo |
| Training Parameters | src/config.py |
| Optimizer | Adam |
| Loss Function | SoftmaxCrossEntropyWithLogits |
| outputs | |
| Loss | 0.4234 |
| Accuracy | AUC[0.8055] |
| Total time | 91 min |
| Params (M) | |
| Checkpoint for Fine tuning | |
| Model for inference | |

#### Inference Performance

| Parameters | | |
| -------------------------- | ----------------------------- | ------------------------- |
| Model Version | | |
| Resource | Ascend 910 | Ascend 310 |
| uploaded Date | 05/27/2020 | 05/27/2020 |
| MindSpore Version | 0.2.0 | 0.2.0 |
| Dataset | Criteo | |
| batch_size | 1000 | |
| outputs | | |
| Accuracy | AUC[0.8055] | |
| Speed | | |
| Total time | 35.559s | |
| Model for inference | | |

# ModelZoo Homepage
[Link](https://gitee.com/mindspore/mindspore/tree/master/mindspore/model_zoo)

+ 14
- 0
example/deepfm_criteo/__init__.py View File

@@ -0,0 +1,14 @@
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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
#
# httpwww.apache.orglicensesLICENSE-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.
# ============================================================================

+ 66
- 0
example/deepfm_criteo/eval.py View File

@@ -0,0 +1,66 @@
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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.
# ============================================================================
"""train_criteo."""
import os
import sys
import time
import argparse

from mindspore import context
from mindspore.train.model import Model
from mindspore.train.serialization import load_checkpoint, load_param_into_net

from src.deepfm import ModelBuilder, AUCMetric
from src.config import DataConfig, ModelConfig, TrainConfig
from src.dataset import create_dataset

sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
parser = argparse.ArgumentParser(description='CTR Prediction')
parser.add_argument('--checkpoint_path', type=str, default=None, help='Checkpoint file path')
parser.add_argument('--dataset_path', type=str, default=None, help='Dataset path')

args_opt, _ = parser.parse_known_args()
device_id = int(os.getenv('DEVICE_ID'))
context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=device_id)


def add_write(file_path, print_str):
with open(file_path, 'a+', encoding='utf-8') as file_out:
file_out.write(print_str + '\n')


if __name__ == '__main__':
data_config = DataConfig()
model_config = ModelConfig()
train_config = TrainConfig()

ds_eval = create_dataset(args_opt.dataset_path, train_mode=False,
epochs=1, batch_size=train_config.batch_size)
model_builder = ModelBuilder(ModelConfig, TrainConfig)
train_net, eval_net = model_builder.get_train_eval_net()
train_net.set_train()
eval_net.set_train(False)
auc_metric = AUCMetric()
model = Model(train_net, eval_network=eval_net, metrics={"auc": auc_metric})
param_dict = load_checkpoint(args_opt.checkpoint_path)
load_param_into_net(eval_net, param_dict)

start = time.time()
res = model.eval(ds_eval)
eval_time = time.time() - start
time_str = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
out_str = f'{time_str} AUC: {list(res.values())[0]}, eval time: {eval_time}s.'
print(out_str)
add_write('./auc.log', str(out_str))

+ 44
- 0
example/deepfm_criteo/scripts/run_distribute_train.sh View File

@@ -0,0 +1,44 @@
#!/bin/bash
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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.
# ============================================================================
echo "Please run the script as: "
echo "sh scripts/run_distribute_train.sh DEVICE_NUM DATASET_PATH MINDSPORE_HCCL_CONFIG_PAHT"
echo "for example: sh scripts/run_distribute_train.sh 8 /dataset_path /rank_table_8p.json"
echo "After running the script, the network runs in the background, The log will be generated in logx/output.log"


export RANK_SIZE=$1
DATA_URL=$2
export MINDSPORE_HCCL_CONFIG_PAHT=$3

for ((i=0; i<RANK_SIZE;i++))
do
export DEVICE_ID=$i
export RANK_ID=$i
rm -rf log$i
mkdir ./log$i
cp *.py ./log$i
cp -r src ./log$i
cd ./log$i || exit
echo "start training for rank $i, device $DEVICE_ID"
env > env.log
python -u train.py \
--dataset_path=$DATA_URL \
--ckpt_path="checkpoint" \
--eval_file_name='auc.log' \
--loss_file_name='loss.log' \
--do_eval=True > output.log 2>&1 &
cd ../
done

+ 32
- 0
example/deepfm_criteo/scripts/run_eval.sh View File

@@ -0,0 +1,32 @@
#!/bin/bash
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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.
# ============================================================================
echo "Please run the script as: "
echo "sh scripts/run_eval.sh DEVICE_ID DATASET_PATH CHECKPOINT_PATH"
echo "for example: sh scripts/run_eval.sh 0 /dataset_path /checkpoint_path"
echo "After running the script, the network runs in the background, The log will be generated in ms_log/eval_output.log"

export DEVICE_ID=$1
DATA_URL=$2
CHECKPOINT_PATH=$3

mkdir -p ms_log
CUR_DIR=`pwd`
export GLOG_log_dir=${CUR_DIR}/ms_log
export GLOG_logtostderr=0

python -u eval.py \
--dataset_path=$DATA_URL \
--checkpoint_path=$CHECKPOINT_PATH > ms_log/eval_output.log 2>&1 &

+ 34
- 0
example/deepfm_criteo/scripts/run_standalone_train.sh View File

@@ -0,0 +1,34 @@
#!/bin/bash
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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.
# ============================================================================
echo "Please run the script as: "
echo "sh scripts/run_standalone_train.sh DEVICE_ID DATASET_PATH"
echo "for example: sh scripts/run_standalone_train.sh 0 /dataset_path"
echo "After running the script, the network runs in the background, The log will be generated in ms_log/output.log"

export DEVICE_ID=$1
DATA_URL=$2

mkdir -p ms_log
CUR_DIR=`pwd`
export GLOG_log_dir=${CUR_DIR}/ms_log
export GLOG_logtostderr=0

python -u train.py \
--dataset_path=$DATA_URL \
--ckpt_path="checkpoint" \
--eval_file_name='auc.log' \
--loss_file_name='loss.log' \
--do_eval=True > ms_log/output.log 2>&1 &

+ 14
- 0
example/deepfm_criteo/src/__init__.py View File

@@ -0,0 +1,14 @@
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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
#
# httpwww.apache.orglicensesLICENSE-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.
# ============================================================================

+ 107
- 0
example/deepfm_criteo/src/callback.py View File

@@ -0,0 +1,107 @@
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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
#
# httpwww.apache.orglicensesLICENSE-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.
# ============================================================================
"""
Defined callback for DeepFM.
"""
import time
from mindspore.train.callback import Callback


def add_write(file_path, out_str):
with open(file_path, 'a+', encoding='utf-8') as file_out:
file_out.write(out_str + '\n')


class EvalCallBack(Callback):
"""
Monitor the loss in training.
If the loss is NAN or INF terminating training.
Note
If per_print_times is 0 do not print loss.
"""
def __init__(self, model, eval_dataset, auc_metric, eval_file_path):
super(EvalCallBack, self).__init__()
self.model = model
self.eval_dataset = eval_dataset
self.aucMetric = auc_metric
self.aucMetric.clear()
self.eval_file_path = eval_file_path

def epoch_end(self, run_context):
start_time = time.time()
out = self.model.eval(self.eval_dataset)
eval_time = int(time.time() - start_time)
time_str = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
out_str = "{} EvalCallBack metric{}; eval_time{}s".format(
time_str, out.values(), eval_time)
print(out_str)
add_write(self.eval_file_path, out_str)


class LossCallBack(Callback):
"""
Monitor the loss in training.
If the loss is NAN or INF terminating training.
Note
If per_print_times is 0 do not print loss.
Args
loss_file_path (str) The file absolute path, to save as loss_file;
per_print_times (int) Print loss every times. Default 1.
"""
def __init__(self, loss_file_path, per_print_times=1):
super(LossCallBack, self).__init__()
if not isinstance(per_print_times, int) or per_print_times < 0:
raise ValueError("print_step must be int and >= 0.")
self.loss_file_path = loss_file_path
self._per_print_times = per_print_times

def step_end(self, run_context):
cb_params = run_context.original_args()
loss = cb_params.net_outputs.asnumpy()
cur_step_in_epoch = (cb_params.cur_step_num - 1) % cb_params.batch_num + 1
cur_num = cb_params.cur_step_num
if self._per_print_times != 0 and cur_num % self._per_print_times == 0:
with open(self.loss_file_path, "a+") as loss_file:
time_str = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
loss_file.write("{} epoch: {} step: {}, loss is {}\n".format(
time_str, cb_params.cur_epoch_num, cur_step_in_epoch, loss))
print("epoch: {} step: {}, loss is {}\n".format(
cb_params.cur_epoch_num, cur_step_in_epoch, loss))


class TimeMonitor(Callback):
"""
Time monitor for calculating cost of each epoch.
Args
data_size (int) step size of an epoch.
"""
def __init__(self, data_size):
super(TimeMonitor, self).__init__()
self.data_size = data_size

def epoch_begin(self, run_context):
self.epoch_time = time.time()

def epoch_end(self, run_context):
epoch_mseconds = (time.time() - self.epoch_time) * 1000
per_step_mseconds = epoch_mseconds / self.data_size
print("epoch time: {0}, per step time: {1}".format(epoch_mseconds, per_step_mseconds), flush=True)

def step_begin(self, run_context):
self.step_time = time.time()

def step_end(self, run_context):
step_mseconds = (time.time() - self.step_time) * 1000
print(f"step time {step_mseconds}", flush=True)

+ 62
- 0
example/deepfm_criteo/src/config.py View File

@@ -0,0 +1,62 @@
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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.
# ============================================================================
"""
network config setting, will be used in train.py and eval.py
"""


class DataConfig:
"""
Define parameters of dataset.
"""
data_vocab_size = 184965
train_num_of_parts = 21
test_num_of_parts = 3
batch_size = 1000
data_field_size = 39
# dataset format, 1: mindrecord, 2: tfrecord, 3: h5
data_format = 2


class ModelConfig:
"""
Define parameters of model.
"""
batch_size = DataConfig.batch_size
data_field_size = DataConfig.data_field_size
data_vocab_size = DataConfig.data_vocab_size
data_emb_dim = 80
deep_layer_args = [[400, 400, 512], "relu"]
init_args = [-0.01, 0.01]
weight_bias_init = ['normal', 'normal']
keep_prob = 0.9


class TrainConfig:
"""
Define parameters of training.
"""
batch_size = DataConfig.batch_size
l2_coef = 1e-6
learning_rate = 1e-5
epsilon = 1e-8
loss_scale = 1024.0
train_epochs = 15
save_checkpoint = True
ckpt_file_name_prefix = "deepfm"
save_checkpoint_steps = 1
keep_checkpoint_max = 15
eval_callback = True
loss_callback = True

+ 299
- 0
example/deepfm_criteo/src/dataset.py View File

@@ -0,0 +1,299 @@
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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.
# ============================================================================
"""
Create train or eval dataset.
"""
import os
import math
from enum import Enum

import pandas as pd
import numpy as np
import mindspore.dataset.engine as de
import mindspore.common.dtype as mstype

from .config import DataConfig


class DataType(Enum):
"""
Enumerate supported dataset format.
"""
MINDRECORD = 1
TFRECORD = 2
H5 = 3


class H5Dataset():
"""
Create dataset with H5 format.

Args:
data_path (str): Dataset directory.
train_mode (bool): Whether dataset is used for train or eval (default=True).
train_num_of_parts (int): The number of train data file (default=21).
test_num_of_parts (int): The number of test data file (default=3).
"""
max_length = 39

def __init__(self, data_path, train_mode=True,
train_num_of_parts=DataConfig.train_num_of_parts,
test_num_of_parts=DataConfig.test_num_of_parts):
self._hdf_data_dir = data_path
self._is_training = train_mode
if self._is_training:
self._file_prefix = 'train'
self._num_of_parts = train_num_of_parts
else:
self._file_prefix = 'test'
self._num_of_parts = test_num_of_parts
self.data_size = self._bin_count(self._hdf_data_dir, self._file_prefix, self._num_of_parts)
print("data_size: {}".format(self.data_size))

def _bin_count(self, hdf_data_dir, file_prefix, num_of_parts):
size = 0
for part in range(num_of_parts):
_y = pd.read_hdf(os.path.join(hdf_data_dir, f'{file_prefix}_output_part_{str(part)}.h5'))
size += _y.shape[0]
return size

def _iterate_hdf_files_(self, num_of_parts=None,
shuffle_block=False):
"""
iterate among hdf files(blocks). when the whole data set is finished, the iterator restarts
from the beginning, thus the data stream will never stop
:param train_mode: True or false,false is eval_mode,
this file iterator will go through the train set
:param num_of_parts: number of files
:param shuffle_block: shuffle block files at every round
:return: input_hdf_file_name, output_hdf_file_name, finish_flag
"""
parts = np.arange(num_of_parts)
while True:
if shuffle_block:
for _ in range(int(shuffle_block)):
np.random.shuffle(parts)
for i, p in enumerate(parts):
yield os.path.join(self._hdf_data_dir, f'{self._file_prefix}_input_part_{str(p)}.h5'), \
os.path.join(self._hdf_data_dir, f'{self._file_prefix}_output_part_{str(p)}.h5'), \
i + 1 == len(parts)

def _generator(self, X, y, batch_size, shuffle=True):
"""
should be accessed only in private
:param X:
:param y:
:param batch_size:
:param shuffle:
:return:
"""
number_of_batches = np.ceil(1. * X.shape[0] / batch_size)
counter = 0
finished = False
sample_index = np.arange(X.shape[0])
if shuffle:
for _ in range(int(shuffle)):
np.random.shuffle(sample_index)
assert X.shape[0] > 0
while True:
batch_index = sample_index[batch_size * counter: batch_size * (counter + 1)]
X_batch = X[batch_index]
y_batch = y[batch_index]
counter += 1
yield X_batch, y_batch, finished
if counter == number_of_batches:
counter = 0
finished = True

def batch_generator(self, batch_size=1000,
random_sample=False, shuffle_block=False):
"""
:param train_mode: True or false,false is eval_mode,
:param batch_size
:param num_of_parts: number of files
:param random_sample: if True, will shuffle
:param shuffle_block: shuffle file blocks at every round
:return:
"""

for hdf_in, hdf_out, _ in self._iterate_hdf_files_(self._num_of_parts,
shuffle_block):
start = stop = None
X_all = pd.read_hdf(hdf_in, start=start, stop=stop).values
y_all = pd.read_hdf(hdf_out, start=start, stop=stop).values
data_gen = self._generator(X_all, y_all, batch_size,
shuffle=random_sample)
finished = False

while not finished:
X, y, finished = data_gen.__next__()
X_id = X[:, 0:self.max_length]
X_va = X[:, self.max_length:]
yield np.array(X_id.astype(dtype=np.int32)), \
np.array(X_va.astype(dtype=np.float32)), \
np.array(y.astype(dtype=np.float32))


def _get_h5_dataset(directory, train_mode=True, epochs=1, batch_size=1000):
"""
Get dataset with h5 format.

Args:
directory (str): Dataset directory.
train_mode (bool): Whether dataset is use for train or eval (default=True).
epochs (int): Dataset epoch size (default=1).
batch_size (int): Dataset batch size (default=1000)

Returns:
Dataset.
"""
data_para = {'batch_size': batch_size}
if train_mode:
data_para['random_sample'] = True
data_para['shuffle_block'] = True

h5_dataset = H5Dataset(data_path=directory, train_mode=train_mode)
numbers_of_batch = math.ceil(h5_dataset.data_size / batch_size)

def _iter_h5_data():
train_eval_gen = h5_dataset.batch_generator(**data_para)
for _ in range(0, numbers_of_batch, 1):
yield train_eval_gen.__next__()

ds = de.GeneratorDataset(_iter_h5_data, ["ids", "weights", "labels"])
ds.set_dataset_size(numbers_of_batch)
ds = ds.repeat(epochs)
return ds


def _get_mindrecord_dataset(directory, train_mode=True, epochs=1, batch_size=1000,
line_per_sample=1000, rank_size=None, rank_id=None):
"""
Get dataset with mindrecord format.

Args:
directory (str): Dataset directory.
train_mode (bool): Whether dataset is use for train or eval (default=True).
epochs (int): Dataset epoch size (default=1).
batch_size (int): Dataset batch size (default=1000).
line_per_sample (int): The number of sample per line (default=1000).
rank_size (int): The number of device, not necessary for single device (default=None).
rank_id (int): Id of device, not necessary for single device (default=None).

Returns:
Dataset.
"""
file_prefix_name = 'train_input_part.mindrecord' if train_mode else 'test_input_part.mindrecord'
file_suffix_name = '00' if train_mode else '0'
shuffle = train_mode

if rank_size is not None and rank_id is not None:
ds = de.MindDataset(os.path.join(directory, file_prefix_name + file_suffix_name),
columns_list=['feat_ids', 'feat_vals', 'label'],
num_shards=rank_size, shard_id=rank_id, shuffle=shuffle,
num_parallel_workers=8)
else:
ds = de.MindDataset(os.path.join(directory, file_prefix_name + file_suffix_name),
columns_list=['feat_ids', 'feat_vals', 'label'],
shuffle=shuffle, num_parallel_workers=8)
ds = ds.batch(int(batch_size / line_per_sample), drop_remainder=True)
ds = ds.map(operations=(lambda x, y, z: (np.array(x).flatten().reshape(batch_size, 39),
np.array(y).flatten().reshape(batch_size, 39),
np.array(z).flatten().reshape(batch_size, 1))),
input_columns=['feat_ids', 'feat_vals', 'label'],
columns_order=['feat_ids', 'feat_vals', 'label'],
num_parallel_workers=8)
ds = ds.repeat(epochs)
return ds


def _get_tf_dataset(directory, train_mode=True, epochs=1, batch_size=1000,
line_per_sample=1000, rank_size=None, rank_id=None):
"""
Get dataset with tfrecord format.

Args:
directory (str): Dataset directory.
train_mode (bool): Whether dataset is use for train or eval (default=True).
epochs (int): Dataset epoch size (default=1).
batch_size (int): Dataset batch size (default=1000).
line_per_sample (int): The number of sample per line (default=1000).
rank_size (int): The number of device, not necessary for single device (default=None).
rank_id (int): Id of device, not necessary for single device (default=None).

Returns:
Dataset.
"""
dataset_files = []
file_prefixt_name = 'train' if train_mode else 'test'
shuffle = train_mode
for (dir_path, _, filenames) in os.walk(directory):
for filename in filenames:
if file_prefixt_name in filename and 'tfrecord' in filename:
dataset_files.append(os.path.join(dir_path, filename))
schema = de.Schema()
schema.add_column('feat_ids', de_type=mstype.int32)
schema.add_column('feat_vals', de_type=mstype.float32)
schema.add_column('label', de_type=mstype.float32)
if rank_size is not None and rank_id is not None:
ds = de.TFRecordDataset(dataset_files=dataset_files, shuffle=shuffle,
schema=schema, num_parallel_workers=8,
num_shards=rank_size, shard_id=rank_id,
shard_equal_rows=True)
else:
ds = de.TFRecordDataset(dataset_files=dataset_files, shuffle=shuffle,
schema=schema, num_parallel_workers=8)
ds = ds.batch(int(batch_size / line_per_sample), drop_remainder=True)
ds = ds.map(operations=(lambda x, y, z: (
np.array(x).flatten().reshape(batch_size, 39),
np.array(y).flatten().reshape(batch_size, 39),
np.array(z).flatten().reshape(batch_size, 1))),
input_columns=['feat_ids', 'feat_vals', 'label'],
columns_order=['feat_ids', 'feat_vals', 'label'],
num_parallel_workers=8)
ds = ds.repeat(epochs)
return ds


def create_dataset(directory, train_mode=True, epochs=1, batch_size=1000,
data_type=DataType.TFRECORD, line_per_sample=1000,
rank_size=None, rank_id=None):
"""
Get dataset.

Args:
directory (str): Dataset directory.
train_mode (bool): Whether dataset is use for train or eval (default=True).
epochs (int): Dataset epoch size (default=1).
batch_size (int): Dataset batch size (default=1000).
data_type (DataType): The type of dataset which is one of H5, TFRECORE, MINDRECORD (default=TFRECORD).
line_per_sample (int): The number of sample per line (default=1000).
rank_size (int): The number of device, not necessary for single device (default=None).
rank_id (int): Id of device, not necessary for single device (default=None).

Returns:
Dataset.
"""
if data_type == DataType.MINDRECORD:
return _get_mindrecord_dataset(directory, train_mode, epochs,
batch_size, line_per_sample,
rank_size, rank_id)
if data_type == DataType.TFRECORD:
return _get_tf_dataset(directory, train_mode, epochs, batch_size,
line_per_sample, rank_size=rank_size, rank_id=rank_id)

if rank_size is not None and rank_size > 1:
raise ValueError('Please use mindrecord dataset.')
return _get_h5_dataset(directory, train_mode, epochs, batch_size)

+ 370
- 0
example/deepfm_criteo/src/deepfm.py View File

@@ -0,0 +1,370 @@
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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.
# ============================================================================
""" test_training """
import os

import numpy as np
from sklearn.metrics import roc_auc_score
import mindspore.common.dtype as mstype
from mindspore.ops import functional as F
from mindspore.ops import composite as C
from mindspore.ops import operations as P
from mindspore.nn import Dropout
from mindspore.nn.optim import Adam
from mindspore.nn.metrics import Metric
from mindspore import nn, ParameterTuple, Parameter
from mindspore.common.initializer import Uniform, initializer, Normal
from mindspore.train.callback import ModelCheckpoint, CheckpointConfig

from .callback import EvalCallBack, LossCallBack


np_type = np.float32
ms_type = mstype.float32


class AUCMetric(Metric):
"""AUC metric for DeepFM model."""
def __init__(self):
super(AUCMetric, self).__init__()
self.pred_probs = []
self.true_labels = []

def clear(self):
"""Clear the internal evaluation result."""
self.pred_probs = []
self.true_labels = []

def update(self, *inputs):
batch_predict = inputs[1].asnumpy()
batch_label = inputs[2].asnumpy()
self.pred_probs.extend(batch_predict.flatten().tolist())
self.true_labels.extend(batch_label.flatten().tolist())

def eval(self):
if len(self.true_labels) != len(self.pred_probs):
raise RuntimeError('true_labels.size() is not equal to pred_probs.size()')
auc = roc_auc_score(self.true_labels, self.pred_probs)
return auc


def init_method(method, shape, name, max_val=0.01):
"""
The method of init parameters.

Args:
method (str): The method uses to initialize parameter.
shape (list): The shape of parameter.
name (str): The name of parameter.
max_val (float): Max value in parameter when uses 'random' or 'uniform' to initialize parameter.

Returns:
Parameter.
"""
if method in ['random', 'uniform']:
params = Parameter(initializer(Uniform(max_val), shape, ms_type), name=name)
elif method == "one":
params = Parameter(initializer("ones", shape, ms_type), name=name)
elif method == 'zero':
params = Parameter(initializer("zeros", shape, ms_type), name=name)
elif method == "normal":
params = Parameter(initializer(Normal(max_val), shape, ms_type), name=name)
return params


def init_var_dict(init_args, values):
"""
Init parameter.

Args:
init_args (list): Define max and min value of parameters.
values (list): Define name, shape and init method of parameters.

Returns:
dict, a dict ot Parameter.
"""
var_map = {}
_, _max_val = init_args
for key, shape, init_flag in values:
if key not in var_map.keys():
if init_flag in ['random', 'uniform']:
var_map[key] = Parameter(initializer(Uniform(_max_val), shape, ms_type), name=key)
elif init_flag == "one":
var_map[key] = Parameter(initializer("ones", shape, ms_type), name=key)
elif init_flag == "zero":
var_map[key] = Parameter(initializer("zeros", shape, ms_type), name=key)
elif init_flag == 'normal':
var_map[key] = Parameter(initializer(Normal(_max_val), shape, ms_type), name=key)
return var_map


class DenseLayer(nn.Cell):
"""
Dense Layer for Deep Layer of DeepFM Model;
Containing: activation, matmul, bias_add;
Args:
input_dim (int): the shape of weight at 0-aixs;
output_dim (int): the shape of weight at 1-aixs, and shape of bias
weight_bias_init (list): weight and bias init method, "random", "uniform", "one", "zero", "normal";
act_str (str): activation function method, "relu", "sigmoid", "tanh";
keep_prob (float): Dropout Layer keep_prob_rate;
scale_coef (float): input scale coefficient;
"""
def __init__(self, input_dim, output_dim, weight_bias_init, act_str, keep_prob=0.9, scale_coef=1.0):
super(DenseLayer, self).__init__()
weight_init, bias_init = weight_bias_init
self.weight = init_method(weight_init, [input_dim, output_dim], name="weight")
self.bias = init_method(bias_init, [output_dim], name="bias")
self.act_func = self._init_activation(act_str)
self.matmul = P.MatMul(transpose_b=False)
self.bias_add = P.BiasAdd()
self.cast = P.Cast()
self.dropout = Dropout(keep_prob=keep_prob)
self.mul = P.Mul()
self.realDiv = P.RealDiv()
self.scale_coef = scale_coef

def _init_activation(self, act_str):
act_str = act_str.lower()
if act_str == "relu":
act_func = P.ReLU()
elif act_str == "sigmoid":
act_func = P.Sigmoid()
elif act_str == "tanh":
act_func = P.Tanh()
return act_func

def construct(self, x):
x = self.act_func(x)
if self.training:
x = self.dropout(x)
x = self.mul(x, self.scale_coef)
x = self.cast(x, mstype.float16)
weight = self.cast(self.weight, mstype.float16)
wx = self.matmul(x, weight)
wx = self.cast(wx, mstype.float32)
wx = self.realDiv(wx, self.scale_coef)
output = self.bias_add(wx, self.bias)
return output


class DeepFMModel(nn.Cell):
"""
From paper: "DeepFM: A Factorization-Machine based Neural Network for CTR Prediction"

Args:
batch_size (int): smaple_number of per step in training; (int, batch_size=128)
filed_size (int): input filed number, or called id_feature number; (int, filed_size=39)
vocab_size (int): id_feature vocab size, id dict size; (int, vocab_size=200000)
emb_dim (int): id embedding vector dim, id mapped to embedding vector; (int, emb_dim=100)
deep_layer_args (list): Deep Layer args, layer_dim_list, layer_activator;
(int, deep_layer_args=[[100, 100, 100], "relu"])
init_args (list): init args for Parameter init; (list, init_args=[min, max, seeds])
weight_bias_init (list): weight, bias init method for deep layers;
(list[str], weight_bias_init=['random', 'zero'])
keep_prob (float): if dropout_flag is True, keep_prob rate to keep connect; (float, keep_prob=0.8)
"""
def __init__(self, config):
super(DeepFMModel, self).__init__()

self.batch_size = config.batch_size
self.field_size = config.data_field_size
self.vocab_size = config.data_vocab_size
self.emb_dim = config.data_emb_dim
self.deep_layer_dims_list, self.deep_layer_act = config.deep_layer_args
self.init_args = config.init_args
self.weight_bias_init = config.weight_bias_init
self.keep_prob = config.keep_prob
init_acts = [('W_l2', [self.vocab_size, 1], 'normal'),
('V_l2', [self.vocab_size, self.emb_dim], 'normal'),
('b', [1], 'normal')]
var_map = init_var_dict(self.init_args, init_acts)
self.fm_w = var_map["W_l2"]
self.fm_b = var_map["b"]
self.embedding_table = var_map["V_l2"]
# Deep Layers
self.deep_input_dims = self.field_size * self.emb_dim + 1
self.all_dim_list = [self.deep_input_dims] + self.deep_layer_dims_list + [1]
self.dense_layer_1 = DenseLayer(self.all_dim_list[0], self.all_dim_list[1],
self.weight_bias_init, self.deep_layer_act, self.keep_prob)
self.dense_layer_2 = DenseLayer(self.all_dim_list[1], self.all_dim_list[2],
self.weight_bias_init, self.deep_layer_act, self.keep_prob)
self.dense_layer_3 = DenseLayer(self.all_dim_list[2], self.all_dim_list[3],
self.weight_bias_init, self.deep_layer_act, self.keep_prob)
self.dense_layer_4 = DenseLayer(self.all_dim_list[3], self.all_dim_list[4],
self.weight_bias_init, self.deep_layer_act, self.keep_prob)
# FM, linear Layers
self.Gatherv2 = P.GatherV2()
self.Mul = P.Mul()
self.ReduceSum = P.ReduceSum(keep_dims=False)
self.Reshape = P.Reshape()
self.Square = P.Square()
self.Shape = P.Shape()
self.Tile = P.Tile()
self.Concat = P.Concat(axis=1)
self.Cast = P.Cast()

def construct(self, id_hldr, wt_hldr):
"""
Args:
id_hldr: batch ids; [bs, field_size]
wt_hldr: batch weights; [bs, field_size]
"""

mask = self.Reshape(wt_hldr, (self.batch_size, self.field_size, 1))
# Linear layer
fm_id_weight = self.Gatherv2(self.fm_w, id_hldr, 0)
wx = self.Mul(fm_id_weight, mask)
linear_out = self.ReduceSum(wx, 1)
# FM layer
fm_id_embs = self.Gatherv2(self.embedding_table, id_hldr, 0)
vx = self.Mul(fm_id_embs, mask)
v1 = self.ReduceSum(vx, 1)
v1 = self.Square(v1)
v2 = self.Square(vx)
v2 = self.ReduceSum(v2, 1)
fm_out = 0.5 * self.ReduceSum(v1 - v2, 1)
fm_out = self.Reshape(fm_out, (-1, 1))
# Deep layer
b = self.Reshape(self.fm_b, (1, 1))
b = self.Tile(b, (self.batch_size, 1))
deep_in = self.Reshape(vx, (-1, self.field_size * self.emb_dim))
deep_in = self.Concat((deep_in, b))
deep_in = self.dense_layer_1(deep_in)
deep_in = self.dense_layer_2(deep_in)
deep_in = self.dense_layer_3(deep_in)
deep_out = self.dense_layer_4(deep_in)
out = linear_out + fm_out + deep_out
return out, fm_id_weight, fm_id_embs


class NetWithLossClass(nn.Cell):
"""
NetWithLossClass definition.
"""
def __init__(self, network, l2_coef=1e-6):
super(NetWithLossClass, self).__init__(auto_prefix=False)
self.loss = P.SigmoidCrossEntropyWithLogits()
self.network = network
self.l2_coef = l2_coef
self.Square = P.Square()
self.ReduceMean_false = P.ReduceMean(keep_dims=False)
self.ReduceSum_false = P.ReduceSum(keep_dims=False)

def construct(self, batch_ids, batch_wts, label):
predict, fm_id_weight, fm_id_embs = self.network(batch_ids, batch_wts)
log_loss = self.loss(predict, label)
mean_log_loss = self.ReduceMean_false(log_loss)
l2_loss_w = self.ReduceSum_false(self.Square(fm_id_weight))
l2_loss_v = self.ReduceSum_false(self.Square(fm_id_embs))
l2_loss_all = self.l2_coef * (l2_loss_v + l2_loss_w) * 0.5
loss = mean_log_loss + l2_loss_all
return loss


class TrainStepWrap(nn.Cell):
"""
TrainStepWrap definition
"""
def __init__(self, network, lr=5e-8, eps=1e-8, loss_scale=1000.0):
super(TrainStepWrap, self).__init__(auto_prefix=False)
self.network = network
self.network.set_train()
self.weights = ParameterTuple(network.trainable_params())
self.optimizer = Adam(self.weights, learning_rate=lr, eps=eps, loss_scale=loss_scale)
self.hyper_map = C.HyperMap()
self.grad = C.GradOperation('grad', get_by_list=True, sens_param=True)
self.sens = loss_scale

def construct(self, batch_ids, batch_wts, label):
weights = self.weights
loss = self.network(batch_ids, batch_wts, label)
sens = P.Fill()(P.DType()(loss), P.Shape()(loss), self.sens) #
grads = self.grad(self.network, weights)(batch_ids, batch_wts, label, sens)
return F.depend(loss, self.optimizer(grads))


class PredictWithSigmoid(nn.Cell):
"""
Eval model with sigmoid.
"""
def __init__(self, network):
super(PredictWithSigmoid, self).__init__(auto_prefix=False)
self.network = network
self.sigmoid = P.Sigmoid()

def construct(self, batch_ids, batch_wts, labels):
logits, _, _, = self.network(batch_ids, batch_wts)
pred_probs = self.sigmoid(logits)

return logits, pred_probs, labels


class ModelBuilder:
"""
Model builder for DeepFM.

Args:
model_config (ModelConfig): Model configuration.
train_config (TrainConfig): Train configuration.
"""
def __init__(self, model_config, train_config):
self.model_config = model_config
self.train_config = train_config

def get_callback_list(self, model=None, eval_dataset=None):
"""
Get callbacks which contains checkpoint callback, eval callback and loss callback.

Args:
model (Cell): The network is added callback (default=None).
eval_dataset (Dataset): Dataset for eval (default=None).
"""
callback_list = []
if self.train_config.save_checkpoint:
config_ck = CheckpointConfig(save_checkpoint_steps=self.train_config.save_checkpoint_steps,
keep_checkpoint_max=self.train_config.keep_checkpoint_max)
ckpt_cb = ModelCheckpoint(prefix=self.train_config.ckpt_file_name_prefix,
directory=self.train_config.output_path,
config=config_ck)
callback_list.append(ckpt_cb)
if self.train_config.eval_callback:
if model is None:
raise RuntimeError("train_config.eval_callback is {}; get_callback_list() args model is {}".format(
self.train_config.eval_callback, model))
if eval_dataset is None:
raise RuntimeError("train_config.eval_callback is {}; get_callback_list() "
"args eval_dataset is {}".format(self.train_config.eval_callback, eval_dataset))
auc_metric = AUCMetric()
eval_callback = EvalCallBack(model, eval_dataset, auc_metric,
eval_file_path=os.path.join(self.train_config.output_path,
self.train_config.eval_file_name))
callback_list.append(eval_callback)
if self.train_config.loss_callback:
loss_callback = LossCallBack(loss_file_path=os.path.join(self.train_config.output_path,
self.train_config.loss_file_name))
callback_list.append(loss_callback)
if callback_list:
return callback_list
return None

def get_train_eval_net(self):
deepfm_net = DeepFMModel(self.model_config)
loss_net = NetWithLossClass(deepfm_net, l2_coef=self.train_config.l2_coef)
train_net = TrainStepWrap(loss_net, lr=self.train_config.learning_rate,
eps=self.train_config.epsilon,
loss_scale=self.train_config.loss_scale)
eval_net = PredictWithSigmoid(deepfm_net)
return train_net, eval_net

+ 91
- 0
example/deepfm_criteo/train.py View File

@@ -0,0 +1,91 @@
# Copyright 2020 Huawei Technologies Co., Ltd
#
# 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.
# ============================================================================
"""train_criteo."""
import os
import sys
import argparse

from mindspore import context, ParallelMode
from mindspore.communication.management import init
from mindspore.train.model import Model
from mindspore.train.callback import ModelCheckpoint, CheckpointConfig, TimeMonitor

from src.deepfm import ModelBuilder, AUCMetric
from src.config import DataConfig, ModelConfig, TrainConfig
from src.dataset import create_dataset, DataType
from src.callback import EvalCallBack, LossCallBack

sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
parser = argparse.ArgumentParser(description='CTR Prediction')
parser.add_argument('--dataset_path', type=str, default=None, help='Dataset path')
parser.add_argument('--ckpt_path', type=str, default=None, help='Checkpoint path')
parser.add_argument('--eval_file_name', type=str, default="./auc.log", help='eval file path')
parser.add_argument('--loss_file_name', type=str, default="./loss.log", help='loss file path')
parser.add_argument('--do_eval', type=bool, default=True, help='Do evaluation or not.')

args_opt, _ = parser.parse_known_args()
device_id = int(os.getenv('DEVICE_ID'))
context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", device_id=device_id)


if __name__ == '__main__':
data_config = DataConfig()
model_config = ModelConfig()
train_config = TrainConfig()

rank_size = int(os.environ.get("RANK_SIZE", 1))
if rank_size > 1:
context.reset_auto_parallel_context()
context.set_auto_parallel_context(parallel_mode=ParallelMode.DATA_PARALLEL, mirror_mean=True)
init()
rank_id = int(os.environ.get('RANK_ID'))
else:
rank_size = None
rank_id = None

ds_train = create_dataset(args_opt.dataset_path,
train_mode=True,
epochs=train_config.train_epochs,
batch_size=train_config.batch_size,
data_type=DataType(data_config.data_format),
rank_size=rank_size,
rank_id=rank_id)

model_builder = ModelBuilder(ModelConfig, TrainConfig)
train_net, eval_net = model_builder.get_train_eval_net()
auc_metric = AUCMetric()
model = Model(train_net, eval_network=eval_net, metrics={"auc": auc_metric})

time_callback = TimeMonitor(data_size=ds_train.get_dataset_size())
loss_callback = LossCallBack(loss_file_path=args_opt.loss_file_name)
callback_list = [time_callback, loss_callback]

if train_config.save_checkpoint:
config_ck = CheckpointConfig(save_checkpoint_steps=train_config.save_checkpoint_steps,
keep_checkpoint_max=train_config.keep_checkpoint_max)
ckpt_cb = ModelCheckpoint(prefix=train_config.ckpt_file_name_prefix,
directory=args_opt.ckpt_path,
config=config_ck)
callback_list.append(ckpt_cb)

if args_opt.do_eval:
ds_eval = create_dataset(args_opt.dataset_path, train_mode=False,
epochs=train_config.train_epochs,
batch_size=train_config.batch_size,
data_type=DataType(data_config.data_format))
eval_callback = EvalCallBack(model, ds_eval, auc_metric,
eval_file_path=args_opt.eval_file_name)
callback_list.append(eval_callback)
model.train(train_config.train_epochs, ds_train, callbacks=callback_list)

Write Preview
Loading…
Cancel
Save