restored_model.summary(print_fn=lambda x: log.write(x + '\n'))
#
# train model
#
start_time = datetime.datetime.now()
log.write("start training at {}\n".format(start_time))
csv_logger = CSVLogger(train_file_log, append=True, separator=',')
seqs = pussy.make_vectors(users_train, ratings)
for i in range(epochs_groups):
restored_model.fit_generator(
pussy.train_generator(seqs, users_train, n_movies, left_out, kept),
epochs=predict_every,
steps_per_epoch=steps_per_epoch,
validation_data=pussy.train_generator(seqs, users_train, n_movies,
left_out, kept),
validation_steps=validation_steps,
callbacks=[csv_logger])
restored_model.save("model_restored_file_{}".format(i) + file_parameters)
final_time = datetime.datetime.now()
log.write("\nfinal time {}\n".format(final_time))
log.write("\ntotal training time {}\n\n".format(final_time - start_time))
#
# save model
#
seqs = pussy.make_vectors(users_train, ratings)
start_time = datetime.datetime.now()
log.write( "start training at {}\n".format(start_time) )
csv_logger = CSVLogger(train_file_log, append=True, separator=',')
for i in range(epochs_groups):
model.fit_generator(pussy.train_generator(seqs, users_train, n_movies, left_out, kept), epochs=predict_every, steps_per_epoch=steps_per_epoch,
validation_data=pussy.train_generator(seqs, users_train, n_movies, left_out, kept), validation_steps=validation_steps,
callbacks=[csv_logger])
model.save("model_file_{}".format(i) + file_parameters)
#pussy.evaluate_model(model, users_train, ratings, n_movies).to_csv(path_or_buf="{}_{}_train.csv".format(file_parameters,(1+i)*predict_every), header=True, sep=",", index=False)
#pussy.evaluate_model(model, users_test, ratings, n_movies).to_csv(path_or_buf="{}_{}_test.csv".format(file_parameters,(1+i)*predict_every), header=True, sep=",", index=False)
final_time = datetime.datetime.now()
log.write( "\nfinal time {}\n".format(final_time) )
log.write( "\ntotal training time {}\n\n".format(final_time - start_time) )
model.summary(print_fn=lambda x: log.write(x + '\n'))
#
# train model
#
seqs = pussy.make_vectors(users_train, ratings)
start_time = datetime.datetime.now()
log.write("start training at {}\n".format(start_time))
csv_logger = CSVLogger(train_file_log, append=True, separator=',')
for i in range(epochs_groups):
model.fit_generator(pussy.train_generator(seqs, users_train, n_movies,
left_out, left_out),
epochs=predict_every,
steps_per_epoch=steps_per_epoch,
validation_data=pussy.train_generator(
seqs, users_train, n_movies, left_out, left_out),
validation_steps=validation_steps,
callbacks=[csv_logger])
model.save("model_file_{}".format(i) + file_parameters)
#pussy.evaluate_model(model, users_train, ratings, n_movies).to_csv(path_or_buf="{}_{}_train.csv".format(file_parameters,(1+i)*predict_every), header=True, sep=",", index=False)
#pussy.evaluate_model(model, users_test, ratings, n_movies).to_csv(path_or_buf="{}_{}_test.csv".format(file_parameters,(1+i)*predict_every), header=True, sep=",", index=False)
final_time = datetime.datetime.now()
log.write("\nfinal time {}\n".format(final_time))
log.write("\ntotal training time {}\n\n".format(final_time - start_time))
#