sparse_feat_emb = SparseFeatureMapper(num_of_fields=26,
hash_size=[
1444, 555, 175781, 128509, 306, 19,
11931, 630, 4, 93146, 5161, 174835,
3176, 28, 11255, 165206, 11, 4606,
2017, 4, 172322, 18, 16, 56456, 86,
43356
],
embedding_dim=2)(sparse_input_node)
attention_output = SelfAttentionInteraction()(
[dense_feat_emb, sparse_feat_emb])
bottom_mlp_output = MLPInteraction()([dense_feat_emb])
top_mlp_output = MLPInteraction()([attention_output, bottom_mlp_output])
output = PointWiseOptimizer()(top_mlp_output)
model = CTRRecommender(inputs=[dense_input_node, sparse_input_node],
outputs=output)
# AutoML search and predict.
searcher = Search(
model=model,
tuner='random',
tuner_params={
'max_trials': 2,
'overwrite': True
},
)
searcher.search(x=train_X,
y=train_y,
x_val=val_X,
id_num=10000,
embedding_dim=64)(input)
item_emb_gmf = LatentFactorMapper(feat_column_id=1,
id_num=10000,
embedding_dim=64)(input)
user_emb_mlp = LatentFactorMapper(feat_column_id=0,
id_num=10000,
embedding_dim=64)(input)
item_emb_mlp = LatentFactorMapper(feat_column_id=1,
id_num=10000,
embedding_dim=64)(input)
innerproduct_output = ElementwiseInteraction(elementwise_type="innerporduct")(
[user_emb_gmf, item_emb_gmf])
mlp_output = MLPInteraction()([user_emb_mlp, item_emb_mlp])
output = PointWiseOptimizer()([innerproduct_output, mlp_output])
model = CTRRecommender(inputs=input, outputs=output)
# AutoML search and predict.
searcher = Search(
model=model,
tuner='random',
tuner_params={
'max_trials': 10,
'overwrite': True
},
)
searcher.search(x=train_X,
y=train_y,
x_val=val_X,
y_val=val_y,
# select model
if args.model == 'dlrm':
output = build_dlrm(emb_dict)
if args.model == 'deepfm':
output = build_deepfm(emb_dict)
if args.model == 'crossnet':
output = build_neumf(emb_dict)
if args.model == 'autoint':
output = build_autorec(emb_dict)
if args.model == 'neumf':
output = build_autorec(emb_dict)
if args.model == 'autorec':
output = build_autorec(emb_dict)
output = PointWiseOptimizer()(output)
model = CTRRecommender(inputs=input, outputs=output)
# search and predict.
searcher = Search(
model=model,
tuner=args.search, ## hyperband, bayesian
tuner_params={
'max_trials': args.trials,
'overwrite': True
})
start_time = time.time()
searcher.search(x=train_X,
y=train_y,
x_val=val_X,
y_val=val_y,