TensorFlow Estimator of Deep CTR:DeepFM/NFM/AFM/FNN/PNN(代码 + 实战)

1,915 阅读6分钟
原文链接: zhuanlan.zhihu.com

深度学习在ctr预估领域的应用越来越多,新的模型不断冒出。从ctr预估问题看看f(x)设计—DNN篇整理了各模型之间的联系之后,一直在琢磨这些东西如何在工业界落地。经过几个月的调研,发现目前存在的一些问题:

* 开源的实现基本都是学术界的人在搞,距离工业应用还有较大的鸿沟
* 模型实现大量调用底层API,各版本实现千差万别,代码臃肿难懂,迁移成本较高
* 单机,放到工业场景下跑不动

针对存在的问题做了一些探索,摸索出一套可行方案,有以下特性:

* 读数据采用Dataset API,支持 parallel and prefetch读取
* 通过Estimator model_fn来实现f(x),迁移到其他算法非常方便,只需要改写model_fn f(x)部分
* 支持分布式以及单机多线程训练
* 支持export model,然后用TensorFlow Serving提供线上预测服务

按工业界的套路,完整的机器学习项目应该包含五个部分:特征框架,训练框架,服务框架,评估框架和监控框架,这里只讨论前三个框架。

特征框架 -- logs in,samples out

实验数据集用criteo,特征工程参考: github.com/PaddlePaddl…

#1 连续特征 剔除异常值/归一化
#2 离散特征 剔掉低频,然后统一编码(特征编码需要保存下来,线上预测的时候要用到)

对大规模离散特征建模是用DNN做ctr预估的优势,paper关注点大都放在ID类特征如何做embedding上,至于连续特征如何处理很少讨论,大概有以下3种方式:

--不做embedding
   |1--concat[continuous, emb_vec]做fc
--做embedding
   |2--离散化之后embedding
   |3--类似FM二阶部分, 统一做embedding, <id, val> 离散特征val=1.0

为了模型设计上的简单统一,采用第3种方式,感兴趣的读者可是试试前两种的效果。

训练框架 -- samples in,model out

目前实现了DeepFM/wide_n_deep/NFM/AFM/FNN/PNN几个算法. 以DeepFM为例来看看如何使用TensorFlow Estimator and Datasets API 来实现input_fn and model_fn:

#1 1:0.5 2:0.03519 3:1 4:0.02567 7:0.03708 8:0.01705 9:0.06296 10:0.18185 11:0.02497 12:1 14:0.02565 15:0.03267 17:0.0247 18:0.03158 20:1 22:1 23:0.13169 24:0.02933 27:0.18159 31:0.0177 34:0.02888 38:1 51:1 63:1 132:1 164:1 236:1
def input_fn(filenames, batch_size=32, num_epochs=1, perform_shuffle=False):
    print('Parsing', filenames)
    def decode_libsvm(line):
        columns = tf.string_split([line], ' ')
        labels = tf.string_to_number(columns.values[0], out_type=tf.float32)
        splits = tf.string_split(columns.values[1:], ':')
        id_vals = tf.reshape(splits.values,splits.dense_shape)
        feat_ids, feat_vals = tf.split(id_vals,num_or_size_splits=2,axis=1)
        feat_ids = tf.string_to_number(feat_ids, out_type=tf.int32)
        feat_vals = tf.string_to_number(feat_vals, out_type=tf.float32)
        return {"feat_ids": feat_ids, "feat_vals": feat_vals}, labels

    # Extract lines from input files using the Dataset API, can pass one filename or filename list
    dataset = tf.data.TextLineDataset(filenames).map(decode_libsvm, num_parallel_calls=10).prefetch(500000)    # multi-thread pre-process then prefetch

    # Randomizes input using a window of 256 elements (read into memory)
    if perform_shuffle:
        dataset = dataset.shuffle(buffer_size=256)

    # epochs from blending together.
    dataset = dataset.repeat(num_epochs)
    dataset = dataset.batch(batch_size) # Batch size to use

    iterator = dataset.make_one_shot_iterator()
    batch_features, batch_labels = iterator.get_next()
    return batch_features, batch_labels
def model_fn(features, labels, mode, params):
    """Bulid Model function f(x) for Estimator."""
    #------hyperparameters----
    field_size = params["field_size"]
    feature_size = params["feature_size"]
    embedding_size = params["embedding_size"]
    l2_reg = params["l2_reg"]
    learning_rate = params["learning_rate"]
    layers = map(int, params["deep_layers"].split(','))
    dropout = map(float, params["dropout"].split(','))

    #------bulid weights------
    FM_B = tf.get_variable(name='fm_bias', shape=[1], initializer=tf.constant_initializer(0.0))
    FM_W = tf.get_variable(name='fm_w', shape=[feature_size], initializer=tf.glorot_normal_initializer())
    FM_V = tf.get_variable(name='fm_v', shape=[feature_size, embedding_size], initializer=tf.glorot_normal_initializer())

    #------build feaure-------
    feat_ids  = features['feat_ids']
    feat_ids = tf.reshape(feat_ids,shape=[-1,field_size])
    feat_vals = features['feat_vals']
    feat_vals = tf.reshape(feat_vals,shape=[-1,field_size])

    #------build f(x)------
    with tf.variable_scope("First-order"):
        feat_wgts = tf.nn.embedding_lookup(FM_W, feat_ids) # None * F * 1
        y_w = tf.reduce_sum(tf.multiply(feat_wgts, feat_vals),1)

    with tf.variable_scope("Second-order"):
        embeddings = tf.nn.embedding_lookup(FM_V, feat_ids) # None * F * K
        feat_vals = tf.reshape(feat_vals, shape=[-1, field_size, 1])
        embeddings = tf.multiply(embeddings, feat_vals) #vij*xi
        sum_square = tf.square(tf.reduce_sum(embeddings,1))
        square_sum = tf.reduce_sum(tf.square(embeddings),1)
        y_v = 0.5*tf.reduce_sum(tf.subtract(sum_square, square_sum),1)	# None * 1

    with tf.variable_scope("Deep-part"):
        if FLAGS.batch_norm:
            if mode == tf.estimator.ModeKeys.TRAIN:
                train_phase = True
            else:
                train_phase = False

        deep_inputs = tf.reshape(embeddings,shape=[-1,field_size*embedding_size]) # None * (F*K)
        for i in range(len(layers)):
            #if FLAGS.batch_norm:
            #    deep_inputs = batch_norm_layer(deep_inputs, train_phase=train_phase, scope_bn='bn_%d' %i)
                #normalizer_params.update({'scope': 'bn_%d' %i})
            deep_inputs = tf.contrib.layers.fully_connected(inputs=deep_inputs, num_outputs=layers[i], \
                #normalizer_fn=normalizer_fn, normalizer_params=normalizer_params, \
                weights_regularizer=tf.contrib.layers.l2_regularizer(l2_reg), scope='mlp%d' % i)
            if FLAGS.batch_norm:
                deep_inputs = batch_norm_layer(deep_inputs, train_phase=train_phase, scope_bn='bn_%d' %i)   #放在RELU之后 https://github.com/ducha-aiki/caffenet-benchmark/blob/master/batchnorm.md#bn----before-or-after-relu
            if mode == tf.estimator.ModeKeys.TRAIN:
                deep_inputs = tf.nn.dropout(deep_inputs, keep_prob=dropout[i])                              #Apply Dropout after all BN layers and set dropout=0.8(drop_ratio=0.2)
                #deep_inputs = tf.layers.dropout(inputs=deep_inputs, rate=dropout[i], training=mode == tf.estimator.ModeKeys.TRAIN)

        y_deep = tf.contrib.layers.fully_connected(inputs=deep_inputs, num_outputs=1, activation_fn=tf.identity, \
                weights_regularizer=tf.contrib.layers.l2_regularizer(l2_reg), scope='deep_out')
        y_d = tf.reshape(y_deep,shape=[-1])

    with tf.variable_scope("DeepFM-out"):
        #y_bias = FM_B * tf.ones_like(labels, dtype=tf.float32)  # None * 1  warning;这里不能用label,否则调用predict/export函数会出错,train/evaluate正常;初步判断estimator做了优化,用不到label时不传
        y_bias = FM_B * tf.ones_like(y_d, dtype=tf.float32)     # None * 1
        y = y_bias + y_w + y_v + y_d
        pred = tf.sigmoid(y)

    predictions={"prob": pred}
    export_outputs = {tf.saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY: tf.estimator.export.PredictOutput(predictions)}
    # Provide an estimator spec for `ModeKeys.PREDICT`
    if mode == tf.estimator.ModeKeys.PREDICT:
        return tf.estimator.EstimatorSpec(mode=mode,predictions=predictions,export_outputs=export_outputs)

    #------bulid loss------
    loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=y, labels=labels)) + \
        l2_reg * tf.nn.l2_loss(FM_W) + l2_reg * tf.nn.l2_loss(FM_V)

    # Provide an estimator spec for `ModeKeys.EVAL`
    eval_metric_ops = {
        "auc": tf.metrics.auc(labels, pred)
    }
    if mode == tf.estimator.ModeKeys.EVAL:
        return tf.estimator.EstimatorSpec(mode=mode,predictions=predictions,loss=loss,eval_metric_ops=eval_metric_ops)

    #------bulid optimizer------
    if FLAGS.optimizer == 'Adam':
        optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate, beta1=0.9, beta2=0.999, epsilon=1e-8)
    elif FLAGS.optimizer == 'Adagrad':
        optimizer = tf.train.AdagradOptimizer(learning_rate=learning_rate, initial_accumulator_value=1e-8)
    elif FLAGS.optimizer == 'Momentum':
        optimizer = tf.train.MomentumOptimizer(learning_rate=learning_rate, momentum=0.95)
    elif FLAGS.optimizer == 'ftrl':
        optimizer = tf.train.FtrlOptimizer(learning_rate)

    train_op = optimizer.minimize(loss, global_step=tf.train.get_global_step())

    # Provide an estimator spec for `ModeKeys.TRAIN` modes
    if mode == tf.estimator.ModeKeys.TRAIN:
        return tf.estimator.EstimatorSpec(mode=mode,predictions=predictions,loss=loss,train_op=train_op)

封装成estimator之后,调用非常简单

#train
python DeepFM.py --task_type=train --learning_rate=0.0005 --optimizer=Adam --num_epochs=1 --batch_size=256 --field_size=39 --feature_size=117581 --deep_layers=400,400,400 --dropout=0.5,0.5,0.5 --log_steps=1000 --num_threads=8 --model_dir=./model_ckpt/criteo/DeepFM/ --data_dir=../../data/criteo/

#predict
python DeepFM.py --task_type=infer --learning_rate=0.0005 --optimizer=Adam --num_epochs=1 --batch_size=256 --field_size=39 --feature_size=117581 --deep_layers=400,400,400 --dropout=0.5,0.5,0.5 --log_steps=1000 --num_threads=8 --model_dir=./model_ckpt/criteo/DeepFM/ --data_dir=../../data/criteo/

完整代码: lambdaji/tf_repos

服务框架 -- request in,pctr out

TensorFlow Serving 是一个用于机器学习模型 serving 的高性能开源库。它可以将训练好的机器学习模型部署到线上,使用 gRPC 作为接口接受外部调用。更加让人眼前一亮的是,它支持模型热更新与自动模型版本管理。这意味着一旦部署 TensorFlow Serving 后,你再也不需要为线上服务操心,只需要关心你的线下模型训练。

首先要导出TF-Serving能识别的模型文件

python DeepFM.py --task_type=export --learning_rate=0.0005 --optimizer=Adam --batch_size=256 --field_size=39 --feature_size=117581 --deep_layers=400,400,400 --dropout=0.5,0.5,0.5 --log_steps=1000 --num_threads=8 --model_dir=./model_ckpt/criteo/DeepFM/ --servable_model_dir=./servable_model/

默认以时间戳来管理版本,生成文件如下:

$ ls -lh servable_model/1517971230
|--saved_model.pb
|--variables
  |--variables.data-00000-of-00001
  |--variables.index

然后写一个client发送请求,这里用C++来写

PredictRequest predictRequest;
PredictResponse response;
ClientContext context;

predictRequest.mutable_model_spec()->set_name(model_name);
predictRequest.mutable_model_spec()->set_signature_name(model_signature_name); //serving_default
google::protobuf::Map<tensorflow::string, tensorflow::TensorProto>& inputs = *predictRequest.mutable_inputs();

//feature to tfrequest
std::vector<long>  ids_vec = {1,2,3,4,5,6,7,8,9,10,11,12,13,15,555,1078,17797,26190,26341,28570,35361,35613,
		35984,48424,51364,64053,65964,66206,71628,84088,84119,86889,88280,88283,100288,100300,102447,109932,111823};
std::vector<float> vals_vec = {0.05,0.006633,0.05,0,0.021594,0.008,0.15,0.04,0.362,0.1,0.2,0,0.04,
		1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
tensorflow::TensorProto feat_ids;
for (uint32_t i = 0; i < ids_vec.size(); i++) {
	feat_ids.add_int64_val(ids_vec[i]);
}
feat_ids.mutable_tensor_shape()->add_dim()->set_size(1);	//batch_size
feat_ids.mutable_tensor_shape()->add_dim()->set_size(feat_ids.int64_val_size());
feat_ids.set_dtype(tensorflow::DataType::DT_INT64);
inputs["feat_ids"] = feat_ids;

tensorflow::TensorProto feat_vals;
for (uint32_t i = 0; i < vals_vec.size(); i++) {
	feat_vals.add_float_val(vals_vec[i]);
}
feat_vals.mutable_tensor_shape()->add_dim()->set_size(1);	//batch_size
feat_vals.mutable_tensor_shape()->add_dim()->set_size(feat_vals.float_val_size());	//sample size
feat_vals.set_dtype(tensorflow::DataType::DT_FLOAT);
inputs["feat_vals"] = feat_vals;

Status status = _stub->Predict(&context, predictRequest, &response);

完整代码: lambdaji/tf_repos

生产环境对时耗和性能的要求较高,而DNN的计算量比LR的简单查表操作大得多,往往需要在效果和性能之间做折中. 这个环节比较考验工程能力, 下图是wide_n_deep model放到线上环境的真实数据,可以看到:

截距部分15ms:对应解析请求包,查询redis/tair,转换特征格式以及打log等
斜率部分0.5ms:一条样本forward一次需要的时间

一个比较有意思的现象是:随着进一步放量,平均时耗不升反降,怀疑TF-Serving内部做了cache类的优化.

Model Performance

本来打算调好参再放出来,但是自从把机器跑挂三次就放弃了:(

图上跑出来的效果不好,可能有几个原因:

--特征工程没做好(连续特征不适合做embedding,负采样,shuffle等等)
--模型设计有问题(不确定有没有bug)
--调参,模型没有收敛到一个足够好的解

感兴趣的小伙伴可以fork下来折腾折腾,做人肉层面的并行,比一个人闭门搞快得多.

项目地址:github.com/lambdaji/tf…

最后提前祝大家新年炼丹愉快!

参考资料:

github.com/wnzhang/dee…

github.com/Atomu2014/p…

github.com/hexiangnan/…

github.com/hexiangnan/…

github.com/ChenglongCh…

zhuanlan.zhihu.com/p/32563337

zhuanlan.zhihu.com/p/28202287