def model_fn_dense(features, labels, mode, params):
dense_feature, sparse_feature = build_features()
dense = tf.feature_column.input_layer(features, dense_feature)
sparse = tf.feature_column.input_layer(features, sparse_feature)
field_size = len(dense_feature)
embedding_size = dense_feature[0].variable_shape.as_list()[-1]
embedding_matrix = tf.reshape(
dense,
[-1, field_size, embedding_size]) # batch * field_size *emb_size
with tf.variable_scope('Linear_output'):
linear_output = tf.layers.dense(sparse, units=1)
add_layer_summary('linear_output', linear_output)
with tf.variable_scope('BI_Pooling'):
sum_square = tf.pow(tf.reduce_sum(embedding_matrix, axis=1), 2)
square_sum = tf.reduce_sum(tf.pow(embedding_matrix, 2), axis=1)
dense = tf.subtract(sum_square, square_sum)
add_layer_summary(dense.name, dense)
dense = stack_dense_layer(dense,
params['hidden_units'],
dropout_rate=params['dropout_rate'],
batch_norm=params['batch_norm'],
mode=mode,
add_summary=True)
with tf.variable_scope('output'):
y = linear_output + dense
add_layer_summary('output', y)
return y
def model_fn_dense(features, labels, mode, params):
dense_feature, sparse_feature = build_features()
dense_input = tf.feature_column.input_layer(features, dense_feature)
sparse_input = tf.feature_column.input_layer(features, sparse_feature)
# Linear part
with tf.variable_scope('Linear_component'):
linear_output = tf.layers.dense(sparse_input, units=1)
add_layer_summary('linear_output', linear_output)
# Deep part
dense_output = stack_dense_layer(dense_input,
params['hidden_units'],
params['dropout_rate'],
params['batch_norm'],
mode,
add_summary=True)
# CIN part
emb_size = dense_feature[0].variable_shape.as_list()[-1]
field_size = len(dense_feature)
embedding_matrix = tf.reshape(
dense_input,
[-1, field_size, emb_size]) # batch * field_size * emb_size
add_layer_summary('embedding_matrix', embedding_matrix)
cin_output = cin_layer(embedding_matrix, params['cin_layer_size'],
emb_size, field_size)
with tf.variable_scope('output'):
y = tf.concat([dense_output, cin_output, linear_output], axis=1)
y = tf.layers.dense(y, units=1)
add_layer_summary('output', y)
return y
def model_fn_sparse(features, labels, mode, params):
# hyper parameter
data_params = params['data_params']
field_size = data_params['field_size']
feature_size = data_params['feature_size']
embedding_size = data_params['embedding_size']
# extract feature
feat_ids = tf.reshape(features['feat_ids'],
shape=[-1, field_size]) # batch * field_size
feat_vals = tf.reshape(features['feat_vals'],
shape=[-1, field_size]) # batch * field_size
# extract embedding
with tf.variable_scope('extract_embedding'):
embedding_matrix = sparse_embedding(
feature_size,
embedding_size,
field_size,
feat_ids,
feat_vals,
add_summary=True) # (batch, field_size, embedding_size)
dense_input = tf.reshape(embedding_matrix,
[-1, field_size * embedding_size
]) # (batch, field_size * embedding_size)
# linear part
linear_output = sparse_linear(feature_size,
feat_ids,
feat_vals,
add_summary=True)
# Deep part
dense_output = stack_dense_layer(dense_input,
params['hidden_units'],
params['dropout_rate'],
params['batch_norm'],
mode,
add_summary=True)
# CIN part
cin_output = cin_layer(embedding_matrix, params['cin_layer_size'],
embedding_size, field_size)
# concat and output
with tf.variable_scope('output'):
y = tf.concat([dense_output, cin_output, linear_output], axis=1)
y = tf.layers.dense(y, units=1)
add_layer_summary('output', y)
return y
def model_fn_dense(features, labels, mode, params):
dense_feature, sparse_feature = build_features()
dense_input = tf.feature_column.input_layer(features, dense_feature)
sparse_input = tf.feature_column.input_layer(features, sparse_feature)
# Linear part
with tf.variable_scope('Linear_component'):
linear_output = tf.layers.dense(sparse_input, units=1)
add_layer_summary('linear_output', linear_output)
field_size = len(dense_feature)
emb_size = dense_feature[0].variable_shape.as_list()[-1]
embedding_matrix = tf.reshape(dense_input, [-1, field_size, emb_size])
# SENET_layer to get new embedding matrix
senet_embedding_matrix = SENET_layer(embedding_matrix,
field_size,
emb_size,
pool_op=params['pool_op'],
ratio=params['senet_ratio'])
# combination layer & BI_interaction
BI_org = Bilinear_layer(embedding_matrix,
field_size,
emb_size,
type=params['model_type'],
name='org')
BI_senet = Bilinear_layer(senet_embedding_matrix,
field_size,
emb_size,
type=params['model_type'],
name='senet')
combination_layer = tf.concat([BI_org, BI_senet], axis=1)
# Deep part
dense_output = stack_dense_layer(combination_layer,
params['hidden_units'],
params['dropout_rate'],
params['batch_norm'],
mode,
add_summary=True)
with tf.variable_scope('output'):
y = dense_output + linear_output
add_layer_summary('output', y)
return y
def model_fn_sparse(features, labels, mode, params):
# hyper parameter
data_params = params['data_params']
field_size = data_params['field_size']
feature_size = data_params['feature_size']
embedding_size = data_params['embedding_size']
# extract feature
feat_ids = tf.reshape(features['feat_ids'],
shape=[-1, field_size]) # batch * field_size
feat_vals = tf.reshape(features['feat_vals'],
shape=[-1, field_size]) # batch * field_size
# extract embedding
embedding_matrix = sparse_embedding(feature_size,
embedding_size,
field_size,
feat_ids,
feat_vals,
add_summary=True)
# linear output
linear_output = sparse_linear(feature_size,
feat_ids,
feat_vals,
add_summary=True)
with tf.variable_scope('BI_Pooling'):
sum_square = tf.pow(tf.reduce_sum(embedding_matrix, axis=1), 2)
square_sum = tf.reduce_sum(tf.pow(embedding_matrix, 2), axis=1)
dense = tf.subtract(sum_square, square_sum)
add_layer_summary(dense.name, dense)
# fully connected stacked dense layers
dense = stack_dense_layer(dense,
params['hidden_units'],
dropout_rate=params['dropout_rate'],
batch_norm=params['batch_norm'],
mode=mode,
add_summary=True)
with tf.variable_scope('output'):
y = linear_output + dense
add_layer_summary('output', y)
return y
def model_fn_sparse(features, labels, mode, params):
# hyper parameter
data_params = params['data_params']
field_size = data_params['field_size']
feature_size = data_params['feature_size']
embedding_size = data_params['embedding_size']
# extract feature
feat_ids = tf.reshape(features['feat_ids'],
shape=[-1, field_size]) # (batch, field_size)
feat_vals = tf.reshape(features['feat_vals'],
shape=[-1, field_size]) # (batch, field_size)
# extract embedding
with tf.variable_scope('extract_embedding'):
embedding_matrix = sparse_embedding(
feature_size,
embedding_size,
field_size,
feat_ids,
feat_vals,
add_summary=True) # (batch, field_size, embedding_size)
dense_input = tf.reshape(embedding_matrix,
[-1, field_size * embedding_size
]) # (batch, field_size * embedding_size)
# deep part
dense = stack_dense_layer(dense_input,
params['hidden_units'],
params['dropout_rate'],
params['batch_norm'],
mode,
add_summary=True)
# cross part
xl = cross_layer(dense_input, params['cross_layers'])
with tf.variable_scope('stack'):
x_stack = tf.concat([dense, xl], axis=1)
with tf.variable_scope('output'):
y = tf.layers.dense(x_stack, units=1)
add_layer_summary('output', y)
return y
def model_fn_varlen(features, labels, mode, params):
# ---general embedding layer---
emb_dict = {}
f_dense = build_features(params)
f_dense = tf.compat.v1.feature_column.input_layer(features, f_dense) # 用户嵌入表示 [batch_size, f_num*emb_dim]
emb_dict['dense_emb'] = f_dense
# Embedding Look up: history item list and category list
item_embedding = tf.compat.v1.get_variable(shape = [params['item_count'], params['emb_dim']],
initializer = tf.truncated_normal_initializer(),
name = 'item_embedding')
cate_embedding = tf.compat.v1.get_variable(shape = [params['cate_count'], params['emb_dim']],
initializer = tf.truncated_normal_initializer(),
name = 'cate_embedding')
item_emb = tf.nn.embedding_lookup( item_embedding, features['item'] ) # [batch_size, emb_dim]
emb_dict['item_emb'] = item_emb
item_hist_emb = tf.nn.embedding_lookup( item_embedding, features['hist_item_list'] ) # [batch_size, padded_size, emb_dim]
emb_dict['item_hist_emb'] = item_hist_emb
cate_emb = tf.nn.embedding_lookup( cate_embedding, features['item_cate'] ) # [batch_size, emb_dim]
emb_dict['cate_emb'] = cate_emb
cate_hist_emb = tf.nn.embedding_lookup( cate_embedding, features['hist_cate_list'] ) # [batch_size, padded_size, emb_dim]
emb_dict['cate_hist_emb'] = cate_hist_emb
# ---sequence embedding layer---
seq_pooling_layer(features, params, emb_dict, mode)
target_attention_layer(features, params, emb_dict)
# Concat features
concat_features = []
for f in params['input_features']:
concat_features.append(emb_dict[f])
fc = tf.concat(concat_features, axis=1)
# ---dnn layer---
main_net = moe_layer(fc, params, mode, scope='main_dense_moe')
bias_net = stack_dense_layer(fc, params['hidden_units'], params['dropout_rate'], params['batch_norm'],
mode, scope='bias_dense')
# ---logits layer---
main_y = tf.layers.dense(main_net, units=1, name='main_logit_net')
bias_y = tf.layers.dense(bias_net, units=1, name='bias_logit_net')
return main_y+bias_y
def model_fn_dense(features, labels, mode, params):
dense_feature = build_features()
dense_input = tf.feature_column.input_layer(features, dense_feature)
# deep part
dense = stack_dense_layer(dense_input,
params['hidden_units'],
params['dropout_rate'],
params['batch_norm'],
mode,
add_summary=True)
# cross part
xl = cross_layer(dense_input, params['cross_layers'], params['cross_op'])
with tf.variable_scope('stack'):
x_stack = tf.concat([dense, xl], axis=1)
with tf.variable_scope('output'):
y = tf.layers.dense(x_stack, units=1)
add_layer_summary('output', y)
return y
def model_fn_sparse(features, labels, mode, params):
# hyper parameter
data_params = params['data_params']
field_size = data_params['field_size']
feature_size = data_params['feature_size']
embedding_size = data_params['embedding_size']
# extract feature
feat_ids = tf.reshape(features['feat_ids'],
shape=[-1, field_size]) # batch * field_size
feat_vals = tf.reshape(features['feat_vals'],
shape=[-1, field_size]) # batch * field_size
# extract embedding
with tf.variable_scope('extract_embedding'):
embedding_matrix = sparse_embedding(
feature_size,
embedding_size,
field_size,
feat_ids,
feat_vals,
add_summary=True) # (batch, field_size, embedding_size)
# linear part
linear_output = sparse_linear(feature_size,
feat_ids,
feat_vals,
add_summary=True)
# SENET_layer to get new embedding matrix
senet_embedding_matrix = SENET_layer(embedding_matrix,
field_size,
embedding_size,
pool_op=params['pool_op'],
ratio=params['senet_ratio'])
# combination layer & BI_interaction
BI_org = Bilinear_layer(embedding_matrix,
field_size,
embedding_size,
type=params['model_type'],
name='org')
BI_senet = Bilinear_layer(senet_embedding_matrix,
field_size,
embedding_size,
type=params['model_type'],
name='senet')
combination_layer = tf.concat([BI_org, BI_senet], axis=1)
# Deep part
dense_output = stack_dense_layer(combination_layer,
params['hidden_units'],
params['dropout_rate'],
params['batch_norm'],
mode,
add_summary=True)
with tf.variable_scope('output'):
y = dense_output + linear_output
add_layer_summary('output', y)
return y