test_data.build_negative_samples(data_info, item_gen_mode="random",
seed=2222)
# ========================== retrain begin =============================
model = DeepFM("ranking", data_info, embed_size=16, n_epochs=2,
lr=1e-4, lr_decay=False, reg=None, batch_size=2048,
num_neg=1, use_bn=False, dropout_rate=None,
hidden_units="128,64,32", tf_sess_config=None)
model.rebuild_graph(path="model_path", model_name="deepfm_model",
full_assign=True)
model.fit(train_data, verbose=2, shuffle=True, eval_data=test_data,
metrics=["loss", "balanced_accuracy", "roc_auc", "pr_auc",
"precision", "recall", "map", "ndcg"])
print("prediction: ", model.predict(user=2211, item=110,
feats={"sex": "mmm",
"genre1": "crime"}))
print("recommendation: ", model.recommend_user(
user=2211, n_rec=7, inner_id=False, cold_start="average",
user_feats=pd.Series({"sex": "F", "occupation": 2, "age": 23}),
item_data=all_data.iloc[4:10]))
eval_result = evaluate(model, test, eval_batch_size=8192, k=10,
metrics=["roc_auc", "pr_auc", "precision",
"recall", "map", "ndcg"],
sample_user_num=2048, neg_sample=True,
update_features=False, seed=2222)
print(eval_result)
reg=None,
batch_size=2048,
num_neg=1,
use_bn=False,
dropout_rate=None,
hidden_units="128,64,32",
tf_sess_config=None)
deepfm.fit(train_data,
verbose=2,
shuffle=True,
eval_data=eval_data,
metrics=[
"loss", "balanced_accuracy", "roc_auc", "pr_auc",
"precision", "recall", "map", "ndcg"
])
print("prediction: ", deepfm.predict(user=1, item=2333))
print("recommendation: ", deepfm.recommend_user(user=1, n_rec=7))
reset_state("AutoInt")
autoint = AutoInt("ranking",
data_info,
embed_size=16,
n_epochs=2,
att_embed_size=(8, 8, 8),
num_heads=4,
use_residual=False,
lr=1e-3,
lr_decay=False,
reg=None,
batch_size=2048,
num_neg=1,
dropout_rate=None,
hidden_units="128,64,32",
tf_sess_config=None)
deepfm.fit(train_data,
verbose=2,
shuffle=True,
eval_data=test_data,
metrics=[
"loss", "balanced_accuracy", "roc_auc", "pr_auc",
"precision", "recall", "map", "ndcg"
],
eval_batch_size=8192,
k=10,
sample_user_num=2048)
print("prediction: ", deepfm.predict(user=2211, item=110))
print("recommendation: ", deepfm.recommend_user(user=2211, n_rec=7))
# save data_info, specify model save folder
data_info.save(path="model_path")
# set manual=True will use numpy to save model
# set manual=False will use tf.train.Saver to save model
# set inference=True will only save the necessary variables for prediction and recommendation
deepfm.save(path="model_path",
model_name="deepfm_model",
manual=True,
inference_only=True)
# =========================== load model ==============================
print("\n", "=" * 50, " after load model ", "=" * 50)
# important to reset graph if model is loaded in the same shell.