MCPcopy Create free account
hub / github.com/DeepRec-AI/DeepRec / main

Function main

modelzoo/din/train.py:669–794  ·  view source on GitHub ↗
(tf_config=None, server=None)

Source from the content-addressed store, hash-verified

667
668
669def main(tf_config=None, server=None):
670 # check dataset and count data set size
671 print("Checking dataset...")
672 train_file = args.data_location + '/local_train_splitByUser'
673 test_file = args.data_location + '/local_test_splitByUser'
674 if args.parquet_dataset and not args.tf:
675 train_file += '.parquet'
676 test_file += '.parquet'
677 if (not os.path.exists(train_file)) or (not os.path.exists(test_file)):
678 print("Dataset does not exist in the given data_location.")
679 sys.exit()
680
681 no_of_training_examples = 0
682 no_of_test_examples = 0
683 if args.parquet_dataset and not args.tf:
684 import pyarrow.parquet as pq
685 no_of_training_examples = pq.read_table(train_file).num_rows
686 no_of_test_examples = pq.read_table(test_file).num_rows
687 else:
688 no_of_training_examples = sum(1 for line in open(train_file))
689 no_of_test_examples = sum(1 for line in open(test_file))
690 print("Numbers of training dataset is {}".format(no_of_training_examples))
691 print("Numbers of test dataset is {}".format(no_of_test_examples))
692
693 # set batch size, eporch & steps
694 batch_size = math.ceil(
695 args.batch_size / args.micro_batch
696 ) if args.micro_batch and not args.tf else args.batch_size
697
698 if args.steps == 0:
699 no_of_epochs = 1
700 train_steps = math.ceil(
701 (float(no_of_epochs) * no_of_training_examples) / batch_size)
702 else:
703 no_of_epochs = math.ceil(
704 (float(batch_size) * args.steps) / no_of_training_examples)
705 train_steps = args.steps
706 test_steps = math.ceil(float(no_of_test_examples) / batch_size)
707 print("The training steps is {}".format(train_steps))
708 print("The testing steps is {}".format(test_steps))
709
710 # set fixed random seed
711 tf.set_random_seed(args.seed)
712
713 # set directory path for checkpoint_dir
714 model_dir = os.path.join(args.output_dir,
715 'model_DIN_' + str(int(time.time())))
716 checkpoint_dir = args.checkpoint if args.checkpoint else model_dir
717 print("Saving model checkpoints to " + checkpoint_dir)
718
719 # create data pipline of train & test dataset
720 train_dataset = build_model_input(train_file, batch_size, no_of_epochs)
721 test_dataset = build_model_input(test_file, batch_size, 1)
722
723 dataset_output_types = tf.data.get_output_types(train_dataset)
724 dataset_output_shapes = tf.data.get_output_shapes(test_dataset)
725 iterator = tf.data.Iterator.from_structure(dataset_output_types,
726 dataset_output_shapes)

Callers 1

train.pyFile · 0.70

Calls 14

sumFunction · 0.85
DINClass · 0.85
exitMethod · 0.80
timeMethod · 0.80
from_structureMethod · 0.80
build_model_inputFunction · 0.70
build_feature_columnsFunction · 0.70
trainFunction · 0.70
evalFunction · 0.70
formatMethod · 0.45
joinMethod · 0.45
get_nextMethod · 0.45

Tested by

no test coverage detected