forked from wubero/code2vec
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathpreprocess.sh
More file actions
70 lines (62 loc) · 3.83 KB
/
preprocess.sh
File metadata and controls
70 lines (62 loc) · 3.83 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
#!/usr/bin/env bash
###########################################################
# Change the following values to preprocess a new dataset.
# TRAIN_DIR, VAL_DIR and TEST_DIR should be paths to
# directories containing sub-directories with .java files
# each of {TRAIN_DIR, VAL_DIR and TEST_DIR} should have sub-dirs,
# and data will be extracted from .java files found in those sub-dirs).
# DATASET_NAME is just a name for the currently extracted
# dataset.
# MAX_CONTEXTS is the number of contexts to keep for each
# method (by default 200).
# WORD_VOCAB_SIZE, PATH_VOCAB_SIZE, TARGET_VOCAB_SIZE -
# - the number of words, paths and target words to keep
# in the vocabulary (the top occurring words and paths will be kept).
# The default values are reasonable for a Tesla K80 GPU
# and new
6826
er (12 GB of board memory).
# NUM_THREADS - the number of parallel threads to use. It is
# recommended to use a multi-core machine for the preprocessing
# step and set this value to the number of cores.
# PYTHON - python3 interpreter alias.
DATA_DIR=$1
TRAIN_DIR="${DATA_DIR}/training"
VAL_DIR="${DATA_DIR}/validation"
TEST_DIR="${DATA_DIR}/test"
DATASET_NAME=$2
MAX_CONTEXTS=200
WORD_VOCAB_SIZE=1301136
PATH_VOCAB_SIZE=911417
TARGET_VOCAB_SIZE=261245
NUM_THREADS=64
PYTHON=python3
###########################################################
TRAIN_DATA_FILE=${DATASET_NAME}.train.raw.txt
VAL_DATA_FILE=${DATASET_NAME}.val.raw.txt
TEST_DATA_FILE=${DATASET_NAME}.test.raw.txt
EXTRACTOR_JAR=JavaExtractor/JPredict/target/JavaExtractor-0.0.1-SNAPSHOT.jar
mkdir -p data
mkdir -p data/${DATASET_NAME}
echo "Extracting paths from validation set..."
${PYTHON} JavaExtractor/extract.py --dir ${VAL_DIR} --max_path_length 8 --max_path_width 2 --num_threads ${NUM_THREADS} --jar ${EXTRACTOR_JAR} > ${VAL_DATA_FILE}
echo "Finished extracting paths from validation set"
echo "Extracting paths from test set..."
${PYTHON} JavaExtractor/extract.py --dir ${TEST_DIR} --max_path_length 8 --max_path_width 2 --num_threads ${NUM_THREADS} --jar ${EXTRACTOR_JAR} > ${TEST_DATA_FILE}
echo "Finished extracting paths from test set"
echo "Extracting paths from training set..."
${PYTHON} JavaExtractor/extract.py --dir ${TRAIN_DIR} --max_path_length 8 --max_path_width 2 --num_threads ${NUM_THREADS} --jar ${EXTRACTOR_JAR} | shuf > ${TRAIN_DATA_FILE}
echo "Finished extracting paths from training set"
TARGET_HISTOGRAM_FILE=data/${DATASET_NAME}/${DATASET_NAME}.histo.tgt.c2v
ORIGIN_HISTOGRAM_FILE=data/${DATASET_NAME}/${DATASET_NAME}.histo.ori.c2v
PATH_HISTOGRAM_FILE=data/${DATASET_NAME}/${DATASET_NAME}.histo.path.c2v
echo "Creating histograms from the training data"
cat ${TRAIN_DATA_FILE} | cut -d' ' -f1 | awk '{n[$0]++} END {for (i in n) print i,n[i]}' > ${TARGET_HISTOGRAM_FILE}
cat ${TRAIN_DATA_FILE} | cut -d' ' -f2- | tr ' ' '\n' | cut -d',' -f1,3 | tr ',' '\n' | awk '{n[$0]++} END {for (i in n) print i,n[i]}' > ${ORIGIN_HISTOGRAM_FILE}
cat ${TRAIN_DATA_FILE} | cut -d' ' -f2- | tr ' ' '\n' | cut -d',' -f2 | awk '{n[$0]++} END {for (i in n) print i,n[i]}' > ${PATH_HISTOGRAM_FILE}
${PYTHON} preprocess.py --train_data ${TRAIN_DATA_FILE} --test_data ${TEST_DATA_FILE} --val_data ${VAL_DATA_FILE} \
--max_contexts ${MAX_CONTEXTS} --word_vocab_size ${WORD_VOCAB_SIZE} --path_vocab_size ${PATH_VOCAB_SIZE} \
--target_vocab_size ${TARGET_VOCAB_SIZE} --word_histogram ${ORIGIN_HISTOGRAM_FILE} \
--path_histogram ${PATH_HISTOGRAM_FILE} --target_histogram ${TARGET_HISTOGRAM_FILE} --output_name data/${DATASET_NAME}/${DATASET_NAME}
# If all went well, the raw data files can be deleted, because preprocess.py creates new files
# with truncated and padded number of paths for each example.
rm ${TRAIN_DATA_FILE} ${VAL_DATA_FILE} ${TEST_DATA_FILE} ${TARGET_HISTOGRAM_FILE} ${ORIGIN_HISTOGRAM_FILE} \
${PATH_HISTOGRAM_FILE}