-
Notifications
You must be signed in to change notification settings - Fork 0
/
preprocessing.py
136 lines (102 loc) · 4.95 KB
/
preprocessing.py
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
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
import pprint
import tempfile
import unicodedata
import os
import tensorflow as tf
import tensorflow_transform as tft
# import numpy as np
import apache_beam as beam
from apache_beam.options.pipeline_options import PipelineOptions
import tensorflow_transform.beam as tft_beam
from tensorflow_transform.tf_metadata import dataset_metadata
from tensorflow_transform.tf_metadata import dataset_schema
INPUT_FILENAME = 'data/input/input_data.csv'
OUTPUT_FILENAME = 'data/output/preprocessed_data'
OUTPUT_TRANSFORM_FUNCTION_FOLDER = 'data/output/transform_artfcts'
NUMERIC_FEATURE_KEYS = [
'sepal_length', 'sepal_width', 'petal_length', 'petal_width'
]
LABEL_KEY = 'target'
def create_raw_metadata():
listFeatures = [(name, tf.io.FixedLenFeature([1], tf.float32)) for name in NUMERIC_FEATURE_KEYS] + [
(LABEL_KEY, tf.io.VarLenFeature(tf.string))
]
RAW_DATA_FEATURE_SPEC = dict(listFeatures)
RAW_DATA_METADATA = tft.tf_metadata.dataset_metadata.DatasetMetadata(
tft.tf_metadata.dataset_schema.schema_utils.schema_from_feature_spec(RAW_DATA_FEATURE_SPEC)
)
return RAW_DATA_METADATA
class Split(beam.DoFn):
def __init__(self):
import numpy as np
def process(self, element):
sepal_length, sepal_width, petal_length, petal_width, target = element.split(",")
return [
{
"sepal_length": np.array([float(sepal_length)]),
"sepal_width": np.array([float(sepal_width)]),
"petal_length": np.array([float(petal_length)]),
"petal_width": np.array([float(petal_width)]),
"target": target,
}
]
def preprocess_fn(input_features):
output_features = {}
# Target feature
# This is a SparseTensor because it is optional. Here we fill in a default
# value when it is missing. This is useful when this column is missing during
# inference
sparse = tf.sparse.SparseTensor(indices=input_features[LABEL_KEY].indices,
values=input_features[LABEL_KEY].values,
dense_shape=[input_features[LABEL_KEY].dense_shape[0], 1])
dense = tf.sparse.to_dense(sp_input=sparse, default_value='')
# # Reshaping from a batch of vectors of size 1 to a batch to scalars.
dense = tf.squeeze(dense, axis=1)
dense_integerized = tft.compute_and_apply_vocabulary(dense, vocab_filename="label_index_map")
output_features['target'] = tf.one_hot(dense_integerized, depth=3)
# normalization of continuous variables
output_features['sepal_length_normalized'] = tft.scale_to_z_score(input_features['sepal_length'])
output_features['sepal_width_normalized'] = tft.scale_to_z_score(input_features['sepal_width'])
output_features['petal_length_normalized'] = tft.scale_to_z_score(input_features['petal_length'])
output_features['petal_width_normalized'] = tft.scale_to_z_score(input_features['petal_width'])
return output_features
def analyze_and_transform(raw_dataset, step="Default"):
transformed_dataset, transform_fn = raw_dataset | "{} - Analyze & Transform".format(
step
) >> tft_beam.AnalyzeAndTransformDataset(preprocess_fn)
return transformed_dataset, transform_fn
def write_tfrecords(dataset, location, step="Default"):
transformed_data, transformed_metadata = dataset
(
transformed_data
| "{} - Write Transformed Data".format(step)
>> beam.io.tfrecordio.WriteToTFRecord(
file_path_prefix=os.path.join(location, "{}-".format(step)),
file_name_suffix=".tfrecords",
coder=tft.coders.example_proto_coder.ExampleProtoCoder(transformed_metadata.schema),
)
)
def write_transform_artefacts(transform_fn, location):
(transform_fn | "Write Transform Artifacts" >> tft_beam.tft_beam_io.transform_fn_io.WriteTransformFn(location))
def run_transformation_pipeline(raw_input_location, transformed_data_location,
transform_artefact_location):
pipeline_options = PipelineOptions()
with beam.Pipeline(options=pipeline_options) as pipeline:
with tft_beam.Context(temp_dir=tempfile.mkdtemp()):
raw_data = (
pipeline |
beam.io.ReadFromText(raw_input_location, skip_header_lines=True) |
beam.ParDo(Split())
)
raw_metadata = create_raw_metadata()
raw_dataset = (raw_data, raw_metadata)
transformed_dataset, transform_fn = analyze_and_transform(raw_dataset)
# transformed_dataset[0] | beam.Map(print)
write_tfrecords(transformed_dataset, transformed_data_location)
write_transform_artefacts(transform_fn, transform_artefact_location)
if __name__ == "__main__":
run_transformation_pipeline(
raw_input_location=INPUT_FILENAME,
transformed_data_location=OUTPUT_FILENAME,
transform_artefact_location=OUTPUT_TRANSFORM_FUNCTION_FOLDER
)