[self.mu, self.bu, self.bi, self.p, self.q])
def load(self):
self.mu, self.p, self.q, self.bu, self.bi = np.load(
'latent_factor.npy')
user_names = np.genfromtxt('users_name.csv', dtype=str)
n_users = len(user_names)
item_names = np.genfromtxt('books_ISBN.csv', dtype=str)
n_books = len(item_names)
user_name2id = dict(zip(user_names, range(n_users)))
book_ISBN2id = dict(zip(item_names, range(n_books)))
R_train, _ = my_IO.read_ratings_train(user_name2id,
book_ISBN2id,
implicit=False)
test_user_ids, test_book_ids = my_IO.read_test(user_name2id, book_ISBN2id)
'''
model = Naive(N=n_users, M=n_books)
model.fit(ratings_train)
result = model.predict(test_user_ids, test_book_ids)
np.savetxt('naive_wo_implicit.csv', result.astype(int), fmt='%d')
'''
'''
model = Baseline(N=n_users, M=n_books)
model.fit(ratings_train)
result = model.predict(test_user_ids, test_book_ids)
np.savetxt('baseline_wo_implicit.csv', result.astype(int), fmt='%d')
'''
config.gpu_options.per_process_gpu_memory_fraction = 0.9
user_names = np.genfromtxt('users_name.csv', dtype=str)
user_embeds = my_IO.get_user_embeds('data/users.csv')
user_embeds = np.array([user_embeds[name] for name in user_names])
n_users = len(user_names)
book_ISBNs = np.genfromtxt('books_ISBN.csv', dtype=str)
n_books = len(book_ISBNs)
user_name2id = dict(zip(user_names, range(n_users)))
book_ISBN2id = dict(zip(book_ISBNs, range(n_books)))
split = 0
R_train, R_valid = my_IO.read_ratings_train(user_name2id,
book_ISBN2id,
implicit=False,
split=split)
mu = R_train.sum() / R_train.nnz
with tf.Session(config=config) as sess:
# Initializaing and building model
model = Baseline(
sess=sess,
model_name=FLAGS.model_name,
checkpoint_dirname=FLAGS.checkpoint_dirname,
)
model.build(mu=mu, N=n_users, M=n_books)
model.build_loss()
model.build_optimizer(optimizer='adam', lr=FLAGS.lr)
FLAGS = flags.FLAGS
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
config.gpu_options.per_process_gpu_memory_fraction = 0.9
users_name = np.genfromtxt('users_name.csv', dtype=str)
n_users = len(users_name)
books_ISBN = np.genfromtxt('books_ISBN.csv', dtype=str)
n_books = len(books_ISBN)
users_name2id = dict(zip(users_name, range(n_users)))
books_ISBN2id = dict(zip(books_ISBN, range(n_books)))
R_train, _ = my_IO.read_ratings_train(users_name2id,
books_ISBN2id,
implicit=False,
split=0)
mu = R_train.sum() / R_train.nnz
with tf.Session(config=config) as sess:
# Initializaing and building model
model = Baseline(
sess=sess,
model_name=FLAGS.model_name,
checkpoint_dirname=FLAGS.checkpoint_dirname,
)
model.build(mu=mu, N=n_users, M=n_books)
model.build_loss()
model.build_optimizer(optimizer='adam', lr=FLAGS.lr)