def train_model(self):
self.logger.info(self.evaluator.metrics_info())
self.user_pos_train = csr_to_user_dict_bytime(
self.dataset.time_matrix, self.dataset.train_matrix)
users_list, item_seq_list, item_pos_list = self._generate_sequences()
for epoch in range(self.epochs):
item_neg_list = self._sample_negative(users_list)
data = DataIterator(users_list,
item_seq_list,
item_pos_list,
item_neg_list,
batch_size=self.batch_size,
shuffle=True)
for bat_user, bat_item_seq, bat_item_pos, bat_item_neg in data:
feed = {
self.user_ph: bat_user,
self.item_seq_ph: bat_item_seq,
self.item_pos_ph: bat_item_pos,
self.item_neg_ph: bat_item_neg,
self.is_training: True
}
self.sess.run(self.train_opt, feed_dict=feed)
result = self.evaluate_model()
self.logger.info("epoch %d:\t%s" % (epoch, result))
def __init__(self, sess, dataset, conf):
super(Fossil, self).__init__(dataset, conf)
self.verbose = conf["verbose"]
self.batch_size = conf["batch_size"]
self.num_epochs = conf["epochs"]
self.embedding_size = conf["embedding_size"]
regs = conf["regs"]
self.lambda_bilinear = regs[0]
self.gamma_bilinear = regs[1]
self.reg_eta = regs[2]
self.alpha = conf["alpha"]
self.num_negatives = conf["num_neg"]
self.learning_rate = conf["learning_rate"]
self.learner = conf["learner"]
self.loss_function = conf["loss_function"]
self.is_pairwise = conf["is_pairwise"]
self.high_order = conf["high_order"]
self.num_negatives = conf["num_neg"]
self.init_method = conf["init_method"]
self.stddev = conf["stddev"]
self.num_negatives = conf["num_neg"]
self.num_users = dataset.num_users
self.num_items = dataset.num_items
self.dataset = dataset
self.train_matrix = self.dataset.train_matrix
self.train_dict = csr_to_user_dict_bytime(self.dataset.time_matrix,
self.train_matrix)
self.sess = sess
def __init__(self, sess, dataset, conf):
super(FPMCplus, self).__init__(dataset, conf)
self.learning_rate = conf["learning_rate"]
self.embedding_size = conf["embedding_size"]
self.weight_size = conf["weight_size"]
self.learner = conf["learner"]
self.loss_function = conf["loss_function"]
self.is_pairwise = conf["is_pairwise"]
self.num_epochs = conf["epochs"]
self.reg_mf = conf["reg_mf"]
self.reg_w = conf["reg_w"]
self.batch_size = conf["batch_size"]
self.high_order = conf["high_order"]
self.verbose = conf["verbose"]
self.embed_init_method = conf["embed_init_method"]
self.weight_init_method = conf["weight_init_method"]
self.stddev = float(conf["stddev"])
self.num_negatives = conf["num_neg"]
self.num_users = dataset.num_users
self.num_items = dataset.num_items
self.dataset = dataset
self.train_matrix = dataset.train_matrix
self.train_dict = csr_to_user_dict_bytime(dataset.time_matrix,
dataset.train_matrix)
self.sess = sess
def get_user_train_dict(self, by_time=False):
if by_time:
train_dict = csr_to_user_dict_bytime(self.time_matrix, self.train_matrix)
else:
train_dict = csr_to_user_dict(self.train_matrix)
return train_dict
def __init__(self, dataset, conf):
super(SASRec, self).__init__(dataset, conf)
train_matrix, time_matrix = dataset.train_matrix, dataset.time_matrix
self.dataset = dataset
self.users_num, self.items_num = train_matrix.shape
self.lr = conf["lr"]
self.l2_emb = conf["l2_emb"]
self.hidden_units = conf["hidden_units"]
self.batch_size = conf["batch_size"]
self.epochs = conf["epochs"]
self.dropout_rate = conf["dropout_rate"]
self.max_len = conf["max_len"]
self.num_blocks = conf["num_blocks"]
self.num_heads = conf["num_heads"]
self.dev = torch.device("cuda")
self.user_pos_train = csr_to_user_dict_bytime(time_matrix,
train_matrix)
self.item_emb = torch.nn.Embedding(self.items_num + 1,
self.hidden_units,
padding_idx=0)
self.pos_emb = torch.nn.Embedding(self.max_len,
self.hidden_units) # TO IMPROVE
self.emb_dropout = torch.nn.Dropout(p=self.dropout_rate)
self.attention_layernorms = torch.nn.ModuleList(
) # to be Q for self-attention
self.attention_layers = torch.nn.ModuleList()
self.forward_layernorms = torch.nn.ModuleList()
self.forward_layers = torch.nn.ModuleList()
self.last_layernorm = torch.nn.LayerNorm(self.hidden_units, eps=1e-8)
for _ in range(self.num_blocks):
new_attn_layernorm = torch.nn.LayerNorm(self.hidden_units,
eps=1e-8)
self.attention_layernorms.append(new_attn_layernorm)
new_attn_layer = torch.nn.MultiheadAttention(
self.hidden_units, self.num_heads, self.dropout_rate)
self.attention_layers.append(new_attn_layer)
new_fwd_layernorm = torch.nn.LayerNorm(self.hidden_units, eps=1e-8)
self.forward_layernorms.append(new_fwd_layernorm)
new_fwd_layer = PointWiseFeedForward(self.hidden_units,
self.dropout_rate)
self.forward_layers.append(new_fwd_layer)
def __init__(self, sess, dataset, conf):
super(NPE, self).__init__(dataset, conf)
self.learning_rate = conf["learning_rate"]
self.embedding_size = conf["embedding_size"]
self.learner = conf["learner"]
self.loss_function = conf["loss_function"]
self.num_epochs = conf["epochs"]
self.reg = conf["reg"]
self.batch_size = conf["batch_size"]
self.high_order = conf["high_order"]
self.verbose = conf["verbose"]
self.num_negatives = conf["num_neg"]
self.init_method = conf["init_method"]
self.stddev = conf["stddev"]
self.num_users = dataset.num_users
self.num_items = dataset.num_items
self.dataset = dataset
self.train_dict = csr_to_user_dict_bytime(dataset.time_matrix,
dataset.train_matrix)
self.sess = sess
def __init__(self, sess, dataset, conf):
super(SASRec, self).__init__(dataset, conf)
train_matrix, time_matrix = dataset.train_matrix, dataset.time_matrix
self.dataset = dataset
self.users_num, self.items_num = train_matrix.shape
self.lr = conf["lr"]
self.l2_reg = conf["l2_reg"]
self.embedding_size = conf["embedding_size"]
self.batch_size = conf["batch_size"]
self.epochs = conf["epochs"]
self.dropout_rate = conf["dropout_rate"]
self.max_len = conf["max_len"]
self.num_blocks = conf["num_blocks"]
self.num_heads = conf["num_heads"]
self.user_pos_train = csr_to_user_dict_bytime(time_matrix,
train_matrix)
self.sess = sess
def __init__(self, sess, dataset, conf):
super(FPMC, self).__init__(dataset, conf)
self.lr = conf["lr"]
self.embedding_size = conf["embedding_size"]
self.learner = conf["learner"]
self.loss_function = conf["loss_function"]
self.is_pairwise = conf["is_pairwise"]
self.topK = conf["topk"]
self.num_epochs = conf["epochs"]
self.reg_mf = conf["reg_mf"]
self.batch_size = conf["batch_size"]
self.init_method = conf["init_method"]
self.stddev = conf["stddev"]
self.verbose = conf["verbose"]
self.num_negatives = conf["num_neg"]
self.num_users = dataset.num_users
self.num_items = dataset.num_items
self.dataset = dataset
self.train_matrix = dataset.train_matrix
self.train_dict = csr_to_user_dict_bytime(dataset.time_matrix,
dataset.train_matrix)
self.sess = sess
