Wprowadzenie poszczególnych komponentów do TensorFlow Extended (TFX)
Zobacz na TensorFlow.org
Uruchom w Google Colab
Wyświetl źródło na GitHub
Pobierz notatnikTen samouczek oparty na Colab interaktywnie przejdzie przez każdy wbudowany komponent TensorFlow Extended (TFX).
Obejmuje każdy etap kompleksowego procesu uczenia maszynowego, od pozyskiwania danych po wypychanie modelu do udostępniania.
Po zakończeniu zawartość tego notesu można automatycznie wyeksportować jako kod źródłowy potoku TFX, który można organizować za pomocą Apache Airflow i Apache Beam.
Tło
Ten notatnik pokazuje, jak używać TFX w środowisku Jupyter/Colab. Tutaj przechodzimy przez przykład Chicago Taxi w interaktywnym notatniku.
Praca w interaktywnym notatniku to przydatny sposób na zapoznanie się ze strukturą potoku TFX. Jest to również przydatne podczas opracowywania własnych potoków jako lekkiego środowiska programistycznego, ale należy pamiętać, że istnieją różnice w sposobie organizowania interaktywnych notesów i uzyskiwaniu przez nie dostępu do artefaktów metadanych.
Orkiestracja
We wdrożeniu produkcyjnym TFX użyjesz koordynatora, takiego jak Apache Airflow, Kubeflow Pipelines lub Apache Beam, aby zaaranżować wstępnie zdefiniowany wykres potoku składników TFX. W interaktywnym notesie sam notes jest koordynatorem, który uruchamia każdy składnik TFX podczas wykonywania komórek notesu.
Metadane
We wdrożeniu produkcyjnym TFX uzyskasz dostęp do metadanych za pośrednictwem interfejsu API ML Metadata (MLMD). MLMD przechowuje właściwości metadanych w bazie danych, takiej jak MySQL lub SQLite, i przechowuje ładunki metadanych w trwałym magazynie, takim jak system plików. W interaktywnym notebooka, obie właściwości i ładunki są przechowywane w bazie danych SQLite efemerycznej w /tmp katalogu na serwerze lub notebooka Jupyter Colab.
Ustawiać
Najpierw instalujemy i importujemy niezbędne pakiety, ustawiamy ścieżki i pobieramy dane.
Ulepsz Pip
Aby uniknąć aktualizacji Pip w systemie uruchomionym lokalnie, upewnij się, że działamy w Colab. Systemy lokalne można oczywiście aktualizować oddzielnie.
try:
import colab
!pip install --upgrade pip
except:
pass
Zainstaluj TFX
pip install -U tfx
Czy uruchomiłeś ponownie środowisko wykonawcze?
Jeśli używasz Google Colab, przy pierwszym uruchomieniu powyższej komórki musisz ponownie uruchomić środowisko wykonawcze (Runtime > Restart runtime ...). Wynika to ze sposobu, w jaki Colab ładuje paczki.
Importuj paczki
Importujemy niezbędne pakiety, w tym standardowe klasy komponentów TFX.
import os
import pprint
import tempfile
import urllib
import absl
import tensorflow as tf
import tensorflow_model_analysis as tfma
tf.get_logger().propagate = False
pp = pprint.PrettyPrinter()
from tfx import v1 as tfx
from tfx.orchestration.experimental.interactive.interactive_context import InteractiveContext
%load_ext tfx.orchestration.experimental.interactive.notebook_extensions.skip
Sprawdźmy wersje bibliotek.
print('TensorFlow version: {}'.format(tf.__version__))
print('TFX version: {}'.format(tfx.__version__))
TensorFlow version: 2.7.0 TFX version: 1.5.0
Skonfiguruj ścieżki potoku
# This is the root directory for your TFX pip package installation.
_tfx_root = tfx.__path__[0]
# This is the directory containing the TFX Chicago Taxi Pipeline example.
_taxi_root = os.path.join(_tfx_root, 'examples/chicago_taxi_pipeline')
# This is the path where your model will be pushed for serving.
_serving_model_dir = os.path.join(
tempfile.mkdtemp(), 'serving_model/taxi_simple')
# Set up logging.
absl.logging.set_verbosity(absl.logging.INFO)
Pobierz przykładowe dane
Pobieramy przykładowy zestaw danych do użycia w naszym potoku TFX.
Zbiór danych używamy jest Taxi Trips zbiór danych wydany przez miasto Chicago. Kolumny w tym zbiorze danych to:
| pickup_community_area | opłata | trip_start_month |
| trip_start_hour | trip_start_day | trip_start_timestamp |
| szerokość_odbioru | długość_odbioru | dropoff_latitude |
| dropoff_longitude | podróż_mile | pickup_census_tract |
| dropoff_census_tract | typ płatności | Spółka |
| trip_seconds | dropoff_community_area | porady |
Ze zbioru danych, będziemy budować model, który przewiduje tips dotyczące podróży.
_data_root = tempfile.mkdtemp(prefix='tfx-data')
DATA_PATH = 'https://raw.githubusercontent.com/tensorflow/tfx/master/tfx/examples/chicago_taxi_pipeline/data/simple/data.csv'
_data_filepath = os.path.join(_data_root, "data.csv")
urllib.request.urlretrieve(DATA_PATH, _data_filepath)
('/tmp/tfx-datacz9xjro6/data.csv', <http.client.HTTPMessage at 0x7f889af49250>)
Rzuć okiem na plik CSV.
head {_data_filepath}
pickup_community_area,fare,trip_start_month,trip_start_hour,trip_start_day,trip_start_timestamp,pickup_latitude,pickup_longitude,dropoff_latitude,dropoff_longitude,trip_miles,pickup_census_tract,dropoff_census_tract,payment_type,company,trip_seconds,dropoff_community_area,tips ,12.45,5,19,6,1400269500,,,,,0.0,,,Credit Card,Chicago Elite Cab Corp. (Chicago Carriag,0,,0.0 ,0,3,19,5,1362683700,,,,,0,,,Unknown,Chicago Elite Cab Corp.,300,,0 60,27.05,10,2,3,1380593700,41.836150155,-87.648787952,,,12.6,,,Cash,Taxi Affiliation Services,1380,,0.0 10,5.85,10,1,2,1382319000,41.985015101,-87.804532006,,,0.0,,,Cash,Taxi Affiliation Services,180,,0.0 14,16.65,5,7,5,1369897200,41.968069,-87.721559063,,,0.0,,,Cash,Dispatch Taxi Affiliation,1080,,0.0 13,16.45,11,12,3,1446554700,41.983636307,-87.723583185,,,6.9,,,Cash,,780,,0.0 16,32.05,12,1,1,1417916700,41.953582125,-87.72345239,,,15.4,,,Cash,,1200,,0.0 30,38.45,10,10,5,1444301100,41.839086906,-87.714003807,,,14.6,,,Cash,,2580,,0.0 11,14.65,1,1,3,1358213400,41.978829526,-87.771166703,,,5.81,,,Cash,,1080,,0.0
Zastrzeżenie: Ta witryna udostępnia aplikacje wykorzystujące dane, które zostały zmodyfikowane do użytku z oryginalnego źródła, www.cityofchicago.org, oficjalnej strony internetowej miasta Chicago. Miasto Chicago nie składa żadnych roszczeń co do treści, dokładności, aktualności ani kompletności jakichkolwiek danych udostępnianych na tej stronie. Dane podane na tej stronie mogą ulec zmianie w dowolnym momencie. Rozumie się, że dane podane na tej stronie są wykorzystywane na własne ryzyko.
Utwórz InteractiveContext
Na koniec tworzymy InteractiveContext, który pozwoli nam na interaktywne uruchamianie komponentów TFX w tym notatniku.
# Here, we create an InteractiveContext using default parameters. This will
# use a temporary directory with an ephemeral ML Metadata database instance.
# To use your own pipeline root or database, the optional properties
# `pipeline_root` and `metadata_connection_config` may be passed to
# InteractiveContext. Calls to InteractiveContext are no-ops outside of the
# notebook.
context = InteractiveContext()
WARNING:absl:InteractiveContext pipeline_root argument not provided: using temporary directory /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq as root for pipeline outputs. WARNING:absl:InteractiveContext metadata_connection_config not provided: using SQLite ML Metadata database at /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/metadata.sqlite.
Interaktywne uruchamianie komponentów TFX
W kolejnych komórkach tworzymy komponenty TFX jeden po drugim, uruchamiamy każdy z nich i wizualizujemy ich artefakty wyjściowe.
Przykład Gen
ExampleGen składnikiem jest zwykle na początku rurociągu TFX. To będzie:
- Podziel dane na zestawy treningowe i ewaluacyjne (domyślnie 2/3 trening + 1/3 ewaluacja)
- Dane konwertowaniu na
tf.Exampleformacie (dowiedz się więcej tutaj ) - Kopiowanie danych do
_tfx_rootkatalogu dla innych komponentów do dostępu
ExampleGen bierze za wejście na ścieżkę do źródła danych. W naszym przypadku jest to _data_root ścieżka, która zawiera pobrane CSV.
example_gen = tfx.components.CsvExampleGen(input_base=_data_root)
context.run(example_gen)
INFO:absl:Running driver for CsvExampleGen INFO:absl:MetadataStore with DB connection initialized INFO:absl:select span and version = (0, None) INFO:absl:latest span and version = (0, None) INFO:absl:Running executor for CsvExampleGen INFO:absl:Generating examples. WARNING:apache_beam.runners.interactive.interactive_environment:Dependencies required for Interactive Beam PCollection visualization are not available, please use: `pip install apache-beam[interactive]` to install necessary dependencies to enable all data visualization features. INFO:absl:Processing input csv data /tmp/tfx-datacz9xjro6/* to TFExample. WARNING:root:Make sure that locally built Python SDK docker image has Python 3.7 interpreter. WARNING:apache_beam.io.tfrecordio:Couldn't find python-snappy so the implementation of _TFRecordUtil._masked_crc32c is not as fast as it could be. INFO:absl:Examples generated. INFO:absl:Running publisher for CsvExampleGen INFO:absl:MetadataStore with DB connection initialized
Przeanalizujmy artefakty wyjściowe ExampleGen . Ten komponent tworzy dwa artefakty, przykłady szkoleniowe i przykłady oceny:
artifact = example_gen.outputs['examples'].get()[0]
print(artifact.split_names, artifact.uri)
["train", "eval"] /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/CsvExampleGen/examples/1
Możemy również przyjrzeć się pierwszym trzem przykładom szkoleń:
# Get the URI of the output artifact representing the training examples, which is a directory
train_uri = os.path.join(example_gen.outputs['examples'].get()[0].uri, 'Split-train')
# Get the list of files in this directory (all compressed TFRecord files)
tfrecord_filenames = [os.path.join(train_uri, name)
for name in os.listdir(train_uri)]
# Create a `TFRecordDataset` to read these files
dataset = tf.data.TFRecordDataset(tfrecord_filenames, compression_type="GZIP")
# Iterate over the first 3 records and decode them.
for tfrecord in dataset.take(3):
serialized_example = tfrecord.numpy()
example = tf.train.Example()
example.ParseFromString(serialized_example)
pp.pprint(example)
features {
feature {
key: "company"
value {
bytes_list {
value: "Chicago Elite Cab Corp. (Chicago Carriag"
}
}
}
feature {
key: "dropoff_census_tract"
value {
int64_list {
}
}
}
feature {
key: "dropoff_community_area"
value {
int64_list {
}
}
}
feature {
key: "dropoff_latitude"
value {
float_list {
}
}
}
feature {
key: "dropoff_longitude"
value {
float_list {
}
}
}
feature {
key: "fare"
value {
float_list {
value: 12.449999809265137
}
}
}
feature {
key: "payment_type"
value {
bytes_list {
value: "Credit Card"
}
}
}
feature {
key: "pickup_census_tract"
value {
int64_list {
}
}
}
feature {
key: "pickup_community_area"
value {
int64_list {
}
}
}
feature {
key: "pickup_latitude"
value {
float_list {
}
}
}
feature {
key: "pickup_longitude"
value {
float_list {
}
}
}
feature {
key: "tips"
value {
float_list {
value: 0.0
}
}
}
feature {
key: "trip_miles"
value {
float_list {
value: 0.0
}
}
}
feature {
key: "trip_seconds"
value {
int64_list {
value: 0
}
}
}
feature {
key: "trip_start_day"
value {
int64_list {
value: 6
}
}
}
feature {
key: "trip_start_hour"
value {
int64_list {
value: 19
}
}
}
feature {
key: "trip_start_month"
value {
int64_list {
value: 5
}
}
}
feature {
key: "trip_start_timestamp"
value {
int64_list {
value: 1400269500
}
}
}
}
features {
feature {
key: "company"
value {
bytes_list {
value: "Taxi Affiliation Services"
}
}
}
feature {
key: "dropoff_census_tract"
value {
int64_list {
}
}
}
feature {
key: "dropoff_community_area"
value {
int64_list {
}
}
}
feature {
key: "dropoff_latitude"
value {
float_list {
}
}
}
feature {
key: "dropoff_longitude"
value {
float_list {
}
}
}
feature {
key: "fare"
value {
float_list {
value: 27.049999237060547
}
}
}
feature {
key: "payment_type"
value {
bytes_list {
value: "Cash"
}
}
}
feature {
key: "pickup_census_tract"
value {
int64_list {
}
}
}
feature {
key: "pickup_community_area"
value {
int64_list {
value: 60
}
}
}
feature {
key: "pickup_latitude"
value {
float_list {
value: 41.836151123046875
}
}
}
feature {
key: "pickup_longitude"
value {
float_list {
value: -87.64878845214844
}
}
}
feature {
key: "tips"
value {
float_list {
value: 0.0
}
}
}
feature {
key: "trip_miles"
value {
float_list {
value: 12.600000381469727
}
}
}
feature {
key: "trip_seconds"
value {
int64_list {
value: 1380
}
}
}
feature {
key: "trip_start_day"
value {
int64_list {
value: 3
}
}
}
feature {
key: "trip_start_hour"
value {
int64_list {
value: 2
}
}
}
feature {
key: "trip_start_month"
value {
int64_list {
value: 10
}
}
}
feature {
key: "trip_start_timestamp"
value {
int64_list {
value: 1380593700
}
}
}
}
features {
feature {
key: "company"
value {
bytes_list {
}
}
}
feature {
key: "dropoff_census_tract"
value {
int64_list {
}
}
}
feature {
key: "dropoff_community_area"
value {
int64_list {
}
}
}
feature {
key: "dropoff_latitude"
value {
float_list {
}
}
}
feature {
key: "dropoff_longitude"
value {
float_list {
}
}
}
feature {
key: "fare"
value {
float_list {
value: 16.450000762939453
}
}
}
feature {
key: "payment_type"
value {
bytes_list {
value: "Cash"
}
}
}
feature {
key: "pickup_census_tract"
value {
int64_list {
}
}
}
feature {
key: "pickup_community_area"
value {
int64_list {
value: 13
}
}
}
feature {
key: "pickup_latitude"
value {
float_list {
value: 41.98363494873047
}
}
}
feature {
key: "pickup_longitude"
value {
float_list {
value: -87.72357940673828
}
}
}
feature {
key: "tips"
value {
float_list {
value: 0.0
}
}
}
feature {
key: "trip_miles"
value {
float_list {
value: 6.900000095367432
}
}
}
feature {
key: "trip_seconds"
value {
int64_list {
value: 780
}
}
}
feature {
key: "trip_start_day"
value {
int64_list {
value: 3
}
}
}
feature {
key: "trip_start_hour"
value {
int64_list {
value: 12
}
}
}
feature {
key: "trip_start_month"
value {
int64_list {
value: 11
}
}
}
feature {
key: "trip_start_timestamp"
value {
int64_list {
value: 1446554700
}
}
}
}
Teraz, ExampleGen zakończył przyjmowaniem danych, następnym krokiem jest analiza danych.
StatystykiGen
W StatisticsGen Oblicza elementów statystyki ponad zbioru danych do analizy danych, jak również do wykorzystania w elementach końcowych. Używa TensorFlow Data Validation bibliotekę.
StatisticsGen bierze jako wejście zestawu danych po prostu spożyty użyciu ExampleGen .
statistics_gen = tfx.components.StatisticsGen(
examples=example_gen.outputs['examples'])
context.run(statistics_gen)
INFO:absl:Excluding no splits because exclude_splits is not set. INFO:absl:Running driver for StatisticsGen INFO:absl:MetadataStore with DB connection initialized INFO:absl:Running executor for StatisticsGen INFO:absl:Generating statistics for split train. INFO:absl:Statistics for split train written to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/StatisticsGen/statistics/2/Split-train. INFO:absl:Generating statistics for split eval. INFO:absl:Statistics for split eval written to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/StatisticsGen/statistics/2/Split-eval. WARNING:root:Make sure that locally built Python SDK docker image has Python 3.7 interpreter. INFO:absl:Running publisher for StatisticsGen INFO:absl:MetadataStore with DB connection initialized
Po StatisticsGen kończy bieg, możemy wizualizować przesyłanych danych statystycznych. Spróbuj pobawić się różnymi fabułami!
context.show(statistics_gen.outputs['statistics'])
SchematGen
SchemaGen komponent generuje schematu opartego na statystykach danych. (Schemat określa oczekiwane granic, typów i właściwości funkcji w swoim zbiorze.) Korzysta on także TensorFlow Data Validation bibliotekę.
SchemaGen weźmie jako wejście statystyki, że generowane z StatisticsGen , patrząc na split szkolenia domyślnie.
schema_gen = tfx.components.SchemaGen(
statistics=statistics_gen.outputs['statistics'],
infer_feature_shape=False)
context.run(schema_gen)
INFO:absl:Excluding no splits because exclude_splits is not set. INFO:absl:Running driver for SchemaGen INFO:absl:MetadataStore with DB connection initialized INFO:absl:Running executor for SchemaGen INFO:absl:Processing schema from statistics for split train. INFO:absl:Processing schema from statistics for split eval. INFO:absl:Schema written to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/SchemaGen/schema/3/schema.pbtxt. INFO:absl:Running publisher for SchemaGen INFO:absl:MetadataStore with DB connection initialized
Po SchemaGen kończy bieg, możemy wizualizować wygenerowany schemat w postaci tabeli.
context.show(schema_gen.outputs['schema'])
Każda funkcja w zestawie danych jest wyświetlana jako wiersz w tabeli schematu wraz z jej właściwościami. Schemat przechwytuje również wszystkie wartości, które przybiera cecha kategorialna, określane jako jej domena.
Aby dowiedzieć się więcej na temat schematów, zobacz dokumentację SchemaGen .
Przykład Validator
ExampleValidator składnik wykrywa anomalie w danych, na podstawie oczekiwań określonych przez schematu. Korzysta również TensorFlow Data Validation bibliotekę.
ExampleValidator weźmie jako wejście statystyki z StatisticsGen , a schemat z SchemaGen .
example_validator = tfx.components.ExampleValidator(
statistics=statistics_gen.outputs['statistics'],
schema=schema_gen.outputs['schema'])
context.run(example_validator)
INFO:absl:Excluding no splits because exclude_splits is not set. INFO:absl:Running driver for ExampleValidator INFO:absl:MetadataStore with DB connection initialized INFO:absl:Running executor for ExampleValidator INFO:absl:Validating schema against the computed statistics for split train. INFO:absl:Validation complete for split train. Anomalies written to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/ExampleValidator/anomalies/4/Split-train. INFO:absl:Validating schema against the computed statistics for split eval. INFO:absl:Validation complete for split eval. Anomalies written to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/ExampleValidator/anomalies/4/Split-eval. INFO:absl:Running publisher for ExampleValidator INFO:absl:MetadataStore with DB connection initialized
Po ExampleValidator kończy bieg, możemy wizualizować anomalie jak stół.
context.show(example_validator.outputs['anomalies'])
W tabeli anomalii widzimy, że anomalii nie ma. Tego byśmy się spodziewali, ponieważ jest to pierwszy zestaw danych, który przeanalizowaliśmy, a schemat jest do niego dostosowany. Powinieneś przejrzeć ten schemat — wszystko, co nieoczekiwane, oznacza anomalię w danych. Po przejrzeniu schematu można użyć do ochrony przyszłych danych, a wytworzone tutaj anomalie można wykorzystać do debugowania wydajności modelu, zrozumienia, jak dane ewoluują w czasie, i identyfikowania błędów danych.
Przekształcać
Transform wykonuje składowe funkcji inżynierii zarówno szkolenia i serwowania. Używa TensorFlow Transform biblioteka.
Transform weźmie jako wejście dane z ExampleGen , schematu z SchemaGen , jak również moduł, który zawiera zdefiniowane przez użytkownika przekształcić kod.
Zobaczmy przykład zdefiniowanej przez użytkownika Transform poniższy kod (za wprowadzenie do TensorFlow Transform API, zobacz samouczek ). Najpierw definiujemy kilka stałych dla inżynierii funkcji:
_taxi_constants_module_file = 'taxi_constants.py'
%%writefile {_taxi_constants_module_file}
# Categorical features are assumed to each have a maximum value in the dataset.
MAX_CATEGORICAL_FEATURE_VALUES = [24, 31, 12]
CATEGORICAL_FEATURE_KEYS = [
'trip_start_hour', 'trip_start_day', 'trip_start_month',
'pickup_census_tract', 'dropoff_census_tract', 'pickup_community_area',
'dropoff_community_area'
]
DENSE_FLOAT_FEATURE_KEYS = ['trip_miles', 'fare', 'trip_seconds']
# Number of buckets used by tf.transform for encoding each feature.
FEATURE_BUCKET_COUNT = 10
BUCKET_FEATURE_KEYS = [
'pickup_latitude', 'pickup_longitude', 'dropoff_latitude',
'dropoff_longitude'
]
# Number of vocabulary terms used for encoding VOCAB_FEATURES by tf.transform
VOCAB_SIZE = 1000
# Count of out-of-vocab buckets in which unrecognized VOCAB_FEATURES are hashed.
OOV_SIZE = 10
VOCAB_FEATURE_KEYS = [
'payment_type',
'company',
]
# Keys
LABEL_KEY = 'tips'
FARE_KEY = 'fare'
Writing taxi_constants.py
Następnie piszemy preprocessing_fn które odbywają się w surowych danych jako dane wejściowe i zwraca przekształconych możliwości, że nasz model może trenować na:
_taxi_transform_module_file = 'taxi_transform.py'
%%writefile {_taxi_transform_module_file}
import tensorflow as tf
import tensorflow_transform as tft
import taxi_constants
_DENSE_FLOAT_FEATURE_KEYS = taxi_constants.DENSE_FLOAT_FEATURE_KEYS
_VOCAB_FEATURE_KEYS = taxi_constants.VOCAB_FEATURE_KEYS
_VOCAB_SIZE = taxi_constants.VOCAB_SIZE
_OOV_SIZE = taxi_constants.OOV_SIZE
_FEATURE_BUCKET_COUNT = taxi_constants.FEATURE_BUCKET_COUNT
_BUCKET_FEATURE_KEYS = taxi_constants.BUCKET_FEATURE_KEYS
_CATEGORICAL_FEATURE_KEYS = taxi_constants.CATEGORICAL_FEATURE_KEYS
_FARE_KEY = taxi_constants.FARE_KEY
_LABEL_KEY = taxi_constants.LABEL_KEY
def preprocessing_fn(inputs):
"""tf.transform's callback function for preprocessing inputs.
Args:
inputs: map from feature keys to raw not-yet-transformed features.
Returns:
Map from string feature key to transformed feature operations.
"""
outputs = {}
for key in _DENSE_FLOAT_FEATURE_KEYS:
# If sparse make it dense, setting nan's to 0 or '', and apply zscore.
outputs[key] = tft.scale_to_z_score(
_fill_in_missing(inputs[key]))
for key in _VOCAB_FEATURE_KEYS:
# Build a vocabulary for this feature.
outputs[key] = tft.compute_and_apply_vocabulary(
_fill_in_missing(inputs[key]),
top_k=_VOCAB_SIZE,
num_oov_buckets=_OOV_SIZE)
for key in _BUCKET_FEATURE_KEYS:
outputs[key] = tft.bucketize(
_fill_in_missing(inputs[key]), _FEATURE_BUCKET_COUNT)
for key in _CATEGORICAL_FEATURE_KEYS:
outputs[key] = _fill_in_missing(inputs[key])
# Was this passenger a big tipper?
taxi_fare = _fill_in_missing(inputs[_FARE_KEY])
tips = _fill_in_missing(inputs[_LABEL_KEY])
outputs[_LABEL_KEY] = tf.where(
tf.math.is_nan(taxi_fare),
tf.cast(tf.zeros_like(taxi_fare), tf.int64),
# Test if the tip was > 20% of the fare.
tf.cast(
tf.greater(tips, tf.multiply(taxi_fare, tf.constant(0.2))), tf.int64))
return outputs
def _fill_in_missing(x):
"""Replace missing values in a SparseTensor.
Fills in missing values of `x` with '' or 0, and converts to a dense tensor.
Args:
x: A `SparseTensor` of rank 2. Its dense shape should have size at most 1
in the second dimension.
Returns:
A rank 1 tensor where missing values of `x` have been filled in.
"""
if not isinstance(x, tf.sparse.SparseTensor):
return x
default_value = '' if x.dtype == tf.string else 0
return tf.squeeze(
tf.sparse.to_dense(
tf.SparseTensor(x.indices, x.values, [x.dense_shape[0], 1]),
default_value),
axis=1)
Writing taxi_transform.py
Teraz przechodzimy w tym kodzie inżynierii funkcja do Transform komponent i uruchomić go przekształcić swoje dane.
transform = tfx.components.Transform(
examples=example_gen.outputs['examples'],
schema=schema_gen.outputs['schema'],
module_file=os.path.abspath(_taxi_transform_module_file))
context.run(transform)
INFO:absl:Generating ephemeral wheel package for '/tmpfs/src/temp/docs/tutorials/tfx/taxi_transform.py' (including modules: ['taxi_transform', 'taxi_constants']).
INFO:absl:User module package has hash fingerprint version f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424.
INFO:absl:Executing: ['/tmpfs/src/tf_docs_env/bin/python', '/tmp/tmp9qnpryw9/_tfx_generated_setup.py', 'bdist_wheel', '--bdist-dir', '/tmp/tmppaskl3va', '--dist-dir', '/tmp/tmpr6oorqji']
/tmpfs/src/tf_docs_env/lib/python3.7/site-packages/setuptools/command/install.py:37: SetuptoolsDeprecationWarning: setup.py install is deprecated. Use build and pip and other standards-based tools.
setuptools.SetuptoolsDeprecationWarning,
INFO:absl:Successfully built user code wheel distribution at '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl'; target user module is 'taxi_transform'.
INFO:absl:Full user module path is 'taxi_transform@/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl'
INFO:absl:Running driver for Transform
INFO:absl:MetadataStore with DB connection initialized
INFO:absl:Running executor for Transform
INFO:absl:Analyze the 'train' split and transform all splits when splits_config is not set.
INFO:absl:udf_utils.get_fn {'module_file': None, 'module_path': 'taxi_transform@/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl', 'preprocessing_fn': None} 'preprocessing_fn'
INFO:absl:Installing '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl' to a temporary directory.
INFO:absl:Executing: ['/tmpfs/src/tf_docs_env/bin/python', '-m', 'pip', 'install', '--target', '/tmp/tmpbvbj9r5b', '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl']
running bdist_wheel
running build
running build_py
creating build
creating build/lib
copying taxi_transform.py -> build/lib
copying taxi_constants.py -> build/lib
running install
running install_lib
running install_egg_info
running egg_info
creating tfx_user_code_Transform.egg-info
writing manifest file 'tfx_user_code_Transform.egg-info/SOURCES.txt'
writing manifest file 'tfx_user_code_Transform.egg-info/SOURCES.txt'
Copying tfx_user_code_Transform.egg-info to /tmp/tmppaskl3va/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3.7.egg-info
running install_scripts
Processing /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl
INFO:absl:Successfully installed '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl'.
INFO:absl:udf_utils.get_fn {'module_file': None, 'module_path': 'taxi_transform@/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl', 'stats_options_updater_fn': None} 'stats_options_updater_fn'
INFO:absl:Installing '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl' to a temporary directory.
INFO:absl:Executing: ['/tmpfs/src/tf_docs_env/bin/python', '-m', 'pip', 'install', '--target', '/tmp/tmpbzwdie1a', '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl']
Installing collected packages: tfx-user-code-Transform
Successfully installed tfx-user-code-Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424
Processing /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl
INFO:absl:Successfully installed '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl'.
INFO:absl:Installing '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl' to a temporary directory.
INFO:absl:Executing: ['/tmpfs/src/tf_docs_env/bin/python', '-m', 'pip', 'install', '--target', '/tmp/tmp09euava5', '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl']
Installing collected packages: tfx-user-code-Transform
Successfully installed tfx-user-code-Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424
Processing /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl
INFO:absl:Successfully installed '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424-py3-none-any.whl'.
INFO:absl:Feature company has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature fare has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature payment_type has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature tips has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_miles has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_seconds has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_day has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_hour has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_month has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_timestamp has no shape. Setting to VarLenSparseTensor.
Installing collected packages: tfx-user-code-Transform
Successfully installed tfx-user-code-Transform-0.0+f78e5f6b4988b5d5289aab277eceaff03bd38343154c2f602e06d95c6acd5424
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.7/site-packages/tensorflow_transform/tf_utils.py:289: Tensor.experimental_ref (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use ref() instead.
INFO:absl:Feature company has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature fare has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature payment_type has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature tips has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_miles has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_seconds has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_day has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_hour has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_month has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_timestamp has no shape. Setting to VarLenSparseTensor.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:Feature company has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature fare has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature payment_type has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature tips has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_miles has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_seconds has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_day has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_hour has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_month has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_timestamp has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature company has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature fare has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature payment_type has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature tips has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_miles has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_seconds has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_day has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_hour has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_month has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_timestamp has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature company has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature fare has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature payment_type has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature tips has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_miles has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_seconds has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_day has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_hour has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_month has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_timestamp has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature company has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature fare has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature payment_type has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature tips has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_miles has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_seconds has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_day has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_hour has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_month has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_timestamp has no shape. Setting to VarLenSparseTensor.
WARNING:root:This output type hint will be ignored and not used for type-checking purposes. Typically, output type hints for a PTransform are single (or nested) types wrapped by a PCollection, PDone, or None. Got: Tuple[Dict[str, Union[NoneType, _Dataset]], Union[Dict[str, Dict[str, PCollection]], NoneType], int] instead.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
WARNING:absl:Tables initialized inside a tf.function will be re-initialized on every invocation of the function. This re-initialization can have significant impact on performance. Consider lifting them out of the graph context using `tf.init_scope`.: compute_and_apply_vocabulary/apply_vocab/text_file_init/InitializeTableFromTextFileV2
WARNING:absl:Tables initialized inside a tf.function will be re-initialized on every invocation of the function. This re-initialization can have significant impact on performance. Consider lifting them out of the graph context using `tf.init_scope`.: compute_and_apply_vocabulary_1/apply_vocab/text_file_init/InitializeTableFromTextFileV2
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
WARNING:absl:Tables initialized inside a tf.function will be re-initialized on every invocation of the function. This re-initialization can have significant impact on performance. Consider lifting them out of the graph context using `tf.init_scope`.: compute_and_apply_vocabulary/apply_vocab/text_file_init/InitializeTableFromTextFileV2
WARNING:absl:Tables initialized inside a tf.function will be re-initialized on every invocation of the function. This re-initialization can have significant impact on performance. Consider lifting them out of the graph context using `tf.init_scope`.: compute_and_apply_vocabulary_1/apply_vocab/text_file_init/InitializeTableFromTextFileV2
WARNING:root:This output type hint will be ignored and not used for type-checking purposes. Typically, output type hints for a PTransform are single (or nested) types wrapped by a PCollection, PDone, or None. Got: Tuple[Dict[str, Union[NoneType, _Dataset]], Union[Dict[str, Dict[str, PCollection]], NoneType], int] instead.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:Feature company has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature fare has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature payment_type has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature tips has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_miles has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_seconds has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_day has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_hour has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_month has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_timestamp has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature company has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature dropoff_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature fare has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature payment_type has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_census_tract has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_community_area has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_latitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature pickup_longitude has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature tips has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_miles has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_seconds has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_day has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_hour has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_month has no shape. Setting to VarLenSparseTensor.
INFO:absl:Feature trip_start_timestamp has no shape. Setting to VarLenSparseTensor.
WARNING:root:Make sure that locally built Python SDK docker image has Python 3.7 interpreter.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
2021-12-21 10:10:18.679569: W tensorflow/python/util/util.cc:368] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.
INFO:tensorflow:Assets written to: /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/transform_graph/5/.temp_path/tftransform_tmp/80dbc09e6ded4a93b5c506e252c8f536/assets
INFO:tensorflow:tensorflow_text is not available.
INFO:tensorflow:tensorflow_decision_forests is not available.
INFO:tensorflow:struct2tensor is not available.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:tensorflow:Assets written to: /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/transform_graph/5/.temp_path/tftransform_tmp/572eacb7c64f4f6e9262f7d496a95f86/assets
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:absl:If the number of unique tokens is smaller than the provided top_k or approximation error is acceptable, consider using tft.experimental.approximate_vocabulary for a potentially more efficient implementation.
INFO:tensorflow:tensorflow_text is not available.
INFO:tensorflow:tensorflow_decision_forests is not available.
INFO:tensorflow:struct2tensor is not available.
INFO:tensorflow:tensorflow_text is not available.
INFO:tensorflow:tensorflow_decision_forests is not available.
INFO:tensorflow:struct2tensor is not available.
INFO:absl:Running publisher for Transform
INFO:absl:MetadataStore with DB connection initialized
Przeanalizujmy artefakty wyjściowe Transform . Ten komponent wytwarza dwa rodzaje wyjść:
-
transform_graphjest wykres, który może wykonywać operacje przerób (ten wykres zostaną uwzględnione w modelach obsługujących i oceny). -
transformed_examplesreprezentuje przetworzonych danych treningowych i oceny.
transform.outputs
{'transform_graph': Channel(
type_name: TransformGraph
artifacts: [Artifact(artifact: id: 5
type_id: 22
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/transform_graph/5"
custom_properties {
key: "name"
value {
string_value: "transform_graph"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Transform"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 22
name: "TransformGraph"
)]
additional_properties: {}
additional_custom_properties: {}
),
'transformed_examples': Channel(
type_name: Examples
artifacts: [Artifact(artifact: id: 6
type_id: 14
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/transformed_examples/5"
properties {
key: "split_names"
value {
string_value: "[\"train\", \"eval\"]"
}
}
custom_properties {
key: "name"
value {
string_value: "transformed_examples"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Transform"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 14
name: "Examples"
properties {
key: "span"
value: INT
}
properties {
key: "split_names"
value: STRING
}
properties {
key: "version"
value: INT
}
base_type: DATASET
)]
additional_properties: {}
additional_custom_properties: {}
),
'updated_analyzer_cache': Channel(
type_name: TransformCache
artifacts: [Artifact(artifact: id: 7
type_id: 23
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/updated_analyzer_cache/5"
custom_properties {
key: "name"
value {
string_value: "updated_analyzer_cache"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Transform"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 23
name: "TransformCache"
)]
additional_properties: {}
additional_custom_properties: {}
),
'pre_transform_schema': Channel(
type_name: Schema
artifacts: [Artifact(artifact: id: 8
type_id: 18
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/pre_transform_schema/5"
custom_properties {
key: "name"
value {
string_value: "pre_transform_schema"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Transform"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 18
name: "Schema"
)]
additional_properties: {}
additional_custom_properties: {}
),
'pre_transform_stats': Channel(
type_name: ExampleStatistics
artifacts: [Artifact(artifact: id: 9
type_id: 16
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/pre_transform_stats/5"
custom_properties {
key: "name"
value {
string_value: "pre_transform_stats"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Transform"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 16
name: "ExampleStatistics"
properties {
key: "span"
value: INT
}
properties {
key: "split_names"
value: STRING
}
base_type: STATISTICS
)]
additional_properties: {}
additional_custom_properties: {}
),
'post_transform_schema': Channel(
type_name: Schema
artifacts: [Artifact(artifact: id: 10
type_id: 18
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/post_transform_schema/5"
custom_properties {
key: "name"
value {
string_value: "post_transform_schema"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Transform"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 18
name: "Schema"
)]
additional_properties: {}
additional_custom_properties: {}
),
'post_transform_stats': Channel(
type_name: ExampleStatistics
artifacts: [Artifact(artifact: id: 11
type_id: 16
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/post_transform_stats/5"
custom_properties {
key: "name"
value {
string_value: "post_transform_stats"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Transform"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 16
name: "ExampleStatistics"
properties {
key: "span"
value: INT
}
properties {
key: "split_names"
value: STRING
}
base_type: STATISTICS
)]
additional_properties: {}
additional_custom_properties: {}
),
'post_transform_anomalies': Channel(
type_name: ExampleAnomalies
artifacts: [Artifact(artifact: id: 12
type_id: 20
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Transform/post_transform_anomalies/5"
custom_properties {
key: "name"
value {
string_value: "post_transform_anomalies"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Transform"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 20
name: "ExampleAnomalies"
properties {
key: "span"
value: INT
}
properties {
key: "split_names"
value: STRING
}
)]
additional_properties: {}
additional_custom_properties: {}
)}
Rzuć okiem na transform_graph artefaktu. Wskazuje na katalog zawierający trzy podkatalogi.
train_uri = transform.outputs['transform_graph'].get()[0].uri
os.listdir(train_uri)
['transform_fn', 'transformed_metadata', 'metadata']
transformed_metadata podkatalogu zawiera schemat przed obróbką danych. transform_fn podkatalog zawiera rzeczywisty wykres przetwarzania wstępnego. metadata podkatalogu zawiera schemat oryginalnego danych.
Możemy również przyjrzeć się pierwszym trzem przekształconym przykładom:
# Get the URI of the output artifact representing the transformed examples, which is a directory
train_uri = os.path.join(transform.outputs['transformed_examples'].get()[0].uri, 'Split-train')
# Get the list of files in this directory (all compressed TFRecord files)
tfrecord_filenames = [os.path.join(train_uri, name)
for name in os.listdir(train_uri)]
# Create a `TFRecordDataset` to read these files
dataset = tf.data.TFRecordDataset(tfrecord_filenames, compression_type="GZIP")
# Iterate over the first 3 records and decode them.
for tfrecord in dataset.take(3):
serialized_example = tfrecord.numpy()
example = tf.train.Example()
example.ParseFromString(serialized_example)
pp.pprint(example)
features {
feature {
key: "company"
value {
int64_list {
value: 8
}
}
}
feature {
key: "dropoff_census_tract"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_community_area"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_latitude"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_longitude"
value {
int64_list {
value: 9
}
}
}
feature {
key: "fare"
value {
float_list {
value: 0.061060599982738495
}
}
}
feature {
key: "payment_type"
value {
int64_list {
value: 1
}
}
}
feature {
key: "pickup_census_tract"
value {
int64_list {
value: 0
}
}
}
feature {
key: "pickup_community_area"
value {
int64_list {
value: 0
}
}
}
feature {
key: "pickup_latitude"
value {
int64_list {
value: 0
}
}
}
feature {
key: "pickup_longitude"
value {
int64_list {
value: 9
}
}
}
feature {
key: "tips"
value {
int64_list {
value: 0
}
}
}
feature {
key: "trip_miles"
value {
float_list {
value: -0.15886741876602173
}
}
}
feature {
key: "trip_seconds"
value {
float_list {
value: -0.7118487358093262
}
}
}
feature {
key: "trip_start_day"
value {
int64_list {
value: 6
}
}
}
feature {
key: "trip_start_hour"
value {
int64_list {
value: 19
}
}
}
feature {
key: "trip_start_month"
value {
int64_list {
value: 5
}
}
}
}
features {
feature {
key: "company"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_census_tract"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_community_area"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_latitude"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_longitude"
value {
int64_list {
value: 9
}
}
}
feature {
key: "fare"
value {
float_list {
value: 1.2521240711212158
}
}
}
feature {
key: "payment_type"
value {
int64_list {
value: 0
}
}
}
feature {
key: "pickup_census_tract"
value {
int64_list {
value: 0
}
}
}
feature {
key: "pickup_community_area"
value {
int64_list {
value: 60
}
}
}
feature {
key: "pickup_latitude"
value {
int64_list {
value: 0
}
}
}
feature {
key: "pickup_longitude"
value {
int64_list {
value: 3
}
}
}
feature {
key: "tips"
value {
int64_list {
value: 0
}
}
}
feature {
key: "trip_miles"
value {
float_list {
value: 0.532160758972168
}
}
}
feature {
key: "trip_seconds"
value {
float_list {
value: 0.5509493350982666
}
}
}
feature {
key: "trip_start_day"
value {
int64_list {
value: 3
}
}
}
feature {
key: "trip_start_hour"
value {
int64_list {
value: 2
}
}
}
feature {
key: "trip_start_month"
value {
int64_list {
value: 10
}
}
}
}
features {
feature {
key: "company"
value {
int64_list {
value: 48
}
}
}
feature {
key: "dropoff_census_tract"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_community_area"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_latitude"
value {
int64_list {
value: 0
}
}
}
feature {
key: "dropoff_longitude"
value {
int64_list {
value: 9
}
}
}
feature {
key: "fare"
value {
float_list {
value: 0.3873794376850128
}
}
}
feature {
key: "payment_type"
value {
int64_list {
value: 0
}
}
}
feature {
key: "pickup_census_tract"
value {
int64_list {
value: 0
}
}
}
feature {
key: "pickup_community_area"
value {
int64_list {
value: 13
}
}
}
feature {
key: "pickup_latitude"
value {
int64_list {
value: 9
}
}
}
feature {
key: "pickup_longitude"
value {
int64_list {
value: 0
}
}
}
feature {
key: "tips"
value {
int64_list {
value: 0
}
}
}
feature {
key: "trip_miles"
value {
float_list {
value: 0.21955277025699615
}
}
}
feature {
key: "trip_seconds"
value {
float_list {
value: 0.0019067146349698305
}
}
}
feature {
key: "trip_start_day"
value {
int64_list {
value: 3
}
}
}
feature {
key: "trip_start_hour"
value {
int64_list {
value: 12
}
}
}
feature {
key: "trip_start_month"
value {
int64_list {
value: 11
}
}
}
}
Po Transform komponent przekształciła dane do funkcji, a kolejnym krokiem jest przygotowanie modelu.
Trener
Trainer komponent będzie szkolić model, który można zdefiniować w TensorFlow. Domyślnie support Trainer prognozy API, aby użyć Keras API, trzeba określić Generic Trainer przez setup custom_executor_spec=executor_spec.ExecutorClassSpec(GenericExecutor) w contructor trenera.
Trainer przyjmuje jako dane wejściowe schematu z SchemaGen transformowane dane i wykres z Transform , szkolenia parametrów, a także moduł zawiera zdefiniowany przez użytkownika kodu modelu.
Zobaczmy przykład zdefiniowanej przez użytkownika kodu modelu poniżej (za wprowadzenie do TensorFlow Keras API, zobacz poradnik ):
_taxi_trainer_module_file = 'taxi_trainer.py'
%%writefile {_taxi_trainer_module_file}
from typing import List, Text
import os
from absl import logging
import datetime
import tensorflow as tf
import tensorflow_transform as tft
from tfx import v1 as tfx
from tfx_bsl.public import tfxio
import taxi_constants
_DENSE_FLOAT_FEATURE_KEYS = taxi_constants.DENSE_FLOAT_FEATURE_KEYS
_VOCAB_FEATURE_KEYS = taxi_constants.VOCAB_FEATURE_KEYS
_VOCAB_SIZE = taxi_constants.VOCAB_SIZE
_OOV_SIZE = taxi_constants.OOV_SIZE
_FEATURE_BUCKET_COUNT = taxi_constants.FEATURE_BUCKET_COUNT
_BUCKET_FEATURE_KEYS = taxi_constants.BUCKET_FEATURE_KEYS
_CATEGORICAL_FEATURE_KEYS = taxi_constants.CATEGORICAL_FEATURE_KEYS
_MAX_CATEGORICAL_FEATURE_VALUES = taxi_constants.MAX_CATEGORICAL_FEATURE_VALUES
_LABEL_KEY = taxi_constants.LABEL_KEY
def _get_tf_examples_serving_signature(model, tf_transform_output):
"""Returns a serving signature that accepts `tensorflow.Example`."""
# We need to track the layers in the model in order to save it.
# TODO(b/162357359): Revise once the bug is resolved.
model.tft_layer_inference = tf_transform_output.transform_features_layer()
@tf.function(input_signature=[
tf.TensorSpec(shape=[None], dtype=tf.string, name='examples')
])
def serve_tf_examples_fn(serialized_tf_example):
"""Returns the output to be used in the serving signature."""
raw_feature_spec = tf_transform_output.raw_feature_spec()
# Remove label feature since these will not be present at serving time.
raw_feature_spec.pop(_LABEL_KEY)
raw_features = tf.io.parse_example(serialized_tf_example, raw_feature_spec)
transformed_features = model.tft_layer_inference(raw_features)
logging.info('serve_transformed_features = %s', transformed_features)
outputs = model(transformed_features)
# TODO(b/154085620): Convert the predicted labels from the model using a
# reverse-lookup (opposite of transform.py).
return {'outputs': outputs}
return serve_tf_examples_fn
def _get_transform_features_signature(model, tf_transform_output):
"""Returns a serving signature that applies tf.Transform to features."""
# We need to track the layers in the model in order to save it.
# TODO(b/162357359): Revise once the bug is resolved.
model.tft_layer_eval = tf_transform_output.transform_features_layer()
@tf.function(input_signature=[
tf.TensorSpec(shape=[None], dtype=tf.string, name='examples')
])
def transform_features_fn(serialized_tf_example):
"""Returns the transformed_features to be fed as input to evaluator."""
raw_feature_spec = tf_transform_output.raw_feature_spec()
raw_features = tf.io.parse_example(serialized_tf_example, raw_feature_spec)
transformed_features = model.tft_layer_eval(raw_features)
logging.info('eval_transformed_features = %s', transformed_features)
return transformed_features
return transform_features_fn
def _input_fn(file_pattern: List[Text],
data_accessor: tfx.components.DataAccessor,
tf_transform_output: tft.TFTransformOutput,
batch_size: int = 200) -> tf.data.Dataset:
"""Generates features and label for tuning/training.
Args:
file_pattern: List of paths or patterns of input tfrecord files.
data_accessor: DataAccessor for converting input to RecordBatch.
tf_transform_output: A TFTransformOutput.
batch_size: representing the number of consecutive elements of returned
dataset to combine in a single batch
Returns:
A dataset that contains (features, indices) tuple where features is a
dictionary of Tensors, and indices is a single Tensor of label indices.
"""
return data_accessor.tf_dataset_factory(
file_pattern,
tfxio.TensorFlowDatasetOptions(
batch_size=batch_size, label_key=_LABEL_KEY),
tf_transform_output.transformed_metadata.schema)
def _build_keras_model(hidden_units: List[int] = None) -> tf.keras.Model:
"""Creates a DNN Keras model for classifying taxi data.
Args:
hidden_units: [int], the layer sizes of the DNN (input layer first).
Returns:
A keras Model.
"""
real_valued_columns = [
tf.feature_column.numeric_column(key, shape=())
for key in _DENSE_FLOAT_FEATURE_KEYS
]
categorical_columns = [
tf.feature_column.categorical_column_with_identity(
key, num_buckets=_VOCAB_SIZE + _OOV_SIZE, default_value=0)
for key in _VOCAB_FEATURE_KEYS
]
categorical_columns += [
tf.feature_column.categorical_column_with_identity(
key, num_buckets=_FEATURE_BUCKET_COUNT, default_value=0)
for key in _BUCKET_FEATURE_KEYS
]
categorical_columns += [
tf.feature_column.categorical_column_with_identity( # pylint: disable=g-complex-comprehension
key,
num_buckets=num_buckets,
default_value=0) for key, num_buckets in zip(
_CATEGORICAL_FEATURE_KEYS,
_MAX_CATEGORICAL_FEATURE_VALUES)
]
indicator_column = [
tf.feature_column.indicator_column(categorical_column)
for categorical_column in categorical_columns
]
model = _wide_and_deep_classifier(
# TODO(b/139668410) replace with premade wide_and_deep keras model
wide_columns=indicator_column,
deep_columns=real_valued_columns,
dnn_hidden_units=hidden_units or [100, 70, 50, 25])
return model
def _wide_and_deep_classifier(wide_columns, deep_columns, dnn_hidden_units):
"""Build a simple keras wide and deep model.
Args:
wide_columns: Feature columns wrapped in indicator_column for wide (linear)
part of the model.
deep_columns: Feature columns for deep part of the model.
dnn_hidden_units: [int], the layer sizes of the hidden DNN.
Returns:
A Wide and Deep Keras model
"""
# Following values are hard coded for simplicity in this example,
# However prefarably they should be passsed in as hparams.
# Keras needs the feature definitions at compile time.
# TODO(b/139081439): Automate generation of input layers from FeatureColumn.
input_layers = {
colname: tf.keras.layers.Input(name=colname, shape=(), dtype=tf.float32)
for colname in _DENSE_FLOAT_FEATURE_KEYS
}
input_layers.update({
colname: tf.keras.layers.Input(name=colname, shape=(), dtype='int32')
for colname in _VOCAB_FEATURE_KEYS
})
input_layers.update({
colname: tf.keras.layers.Input(name=colname, shape=(), dtype='int32')
for colname in _BUCKET_FEATURE_KEYS
})
input_layers.update({
colname: tf.keras.layers.Input(name=colname, shape=(), dtype='int32')
for colname in _CATEGORICAL_FEATURE_KEYS
})
# TODO(b/161952382): Replace with Keras preprocessing layers.
deep = tf.keras.layers.DenseFeatures(deep_columns)(input_layers)
for numnodes in dnn_hidden_units:
deep = tf.keras.layers.Dense(numnodes)(deep)
wide = tf.keras.layers.DenseFeatures(wide_columns)(input_layers)
output = tf.keras.layers.Dense(1)(
tf.keras.layers.concatenate([deep, wide]))
model = tf.keras.Model(input_layers, output)
model.compile(
loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),
optimizer=tf.keras.optimizers.Adam(lr=0.001),
metrics=[tf.keras.metrics.BinaryAccuracy()])
model.summary(print_fn=logging.info)
return model
# TFX Trainer will call this function.
def run_fn(fn_args: tfx.components.FnArgs):
"""Train the model based on given args.
Args:
fn_args: Holds args used to train the model as name/value pairs.
"""
# Number of nodes in the first layer of the DNN
first_dnn_layer_size = 100
num_dnn_layers = 4
dnn_decay_factor = 0.7
tf_transform_output = tft.TFTransformOutput(fn_args.transform_output)
train_dataset = _input_fn(fn_args.train_files, fn_args.data_accessor,
tf_transform_output, 40)
eval_dataset = _input_fn(fn_args.eval_files, fn_args.data_accessor,
tf_transform_output, 40)
model = _build_keras_model(
# Construct layers sizes with exponetial decay
hidden_units=[
max(2, int(first_dnn_layer_size * dnn_decay_factor**i))
for i in range(num_dnn_layers)
])
tensorboard_callback = tf.keras.callbacks.TensorBoard(
log_dir=fn_args.model_run_dir, update_freq='batch')
model.fit(
train_dataset,
steps_per_epoch=fn_args.train_steps,
validation_data=eval_dataset,
validation_steps=fn_args.eval_steps,
callbacks=[tensorboard_callback])
signatures = {
'serving_default':
_get_tf_examples_serving_signature(model, tf_transform_output),
'transform_features':
_get_transform_features_signature(model, tf_transform_output),
}
model.save(fn_args.serving_model_dir, save_format='tf', signatures=signatures)
Writing taxi_trainer.py
Teraz przechodzimy w tym kodzie modelu do Trainer składnika i uruchomić go trenować model.
trainer = tfx.components.Trainer(
module_file=os.path.abspath(_taxi_trainer_module_file),
examples=transform.outputs['transformed_examples'],
transform_graph=transform.outputs['transform_graph'],
schema=schema_gen.outputs['schema'],
train_args=tfx.proto.TrainArgs(num_steps=10000),
eval_args=tfx.proto.EvalArgs(num_steps=5000))
context.run(trainer)
INFO:absl:Generating ephemeral wheel package for '/tmpfs/src/temp/docs/tutorials/tfx/taxi_trainer.py' (including modules: ['taxi_transform', 'taxi_constants', 'taxi_trainer']).
INFO:absl:User module package has hash fingerprint version ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af.
INFO:absl:Executing: ['/tmpfs/src/tf_docs_env/bin/python', '/tmp/tmpzxd5b1yc/_tfx_generated_setup.py', 'bdist_wheel', '--bdist-dir', '/tmp/tmpbg9ly6tr', '--dist-dir', '/tmp/tmpx43qh690']
/tmpfs/src/tf_docs_env/lib/python3.7/site-packages/setuptools/command/install.py:37: SetuptoolsDeprecationWarning: setup.py install is deprecated. Use build and pip and other standards-based tools.
setuptools.SetuptoolsDeprecationWarning,
INFO:absl:Successfully built user code wheel distribution at '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Trainer-0.0+ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af-py3-none-any.whl'; target user module is 'taxi_trainer'.
INFO:absl:Full user module path is 'taxi_trainer@/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Trainer-0.0+ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af-py3-none-any.whl'
INFO:absl:Running driver for Trainer
INFO:absl:MetadataStore with DB connection initialized
INFO:absl:Running executor for Trainer
INFO:absl:Train on the 'train' split when train_args.splits is not set.
INFO:absl:Evaluate on the 'eval' split when eval_args.splits is not set.
WARNING:absl:Examples artifact does not have payload_format custom property. Falling back to FORMAT_TF_EXAMPLE
WARNING:absl:Examples artifact does not have payload_format custom property. Falling back to FORMAT_TF_EXAMPLE
WARNING:absl:Examples artifact does not have payload_format custom property. Falling back to FORMAT_TF_EXAMPLE
INFO:absl:udf_utils.get_fn {'train_args': '{\n "num_steps": 10000\n}', 'eval_args': '{\n "num_steps": 5000\n}', 'module_file': None, 'run_fn': None, 'trainer_fn': None, 'custom_config': 'null', 'module_path': 'taxi_trainer@/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Trainer-0.0+ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af-py3-none-any.whl'} 'run_fn'
INFO:absl:Installing '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Trainer-0.0+ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af-py3-none-any.whl' to a temporary directory.
INFO:absl:Executing: ['/tmpfs/src/tf_docs_env/bin/python', '-m', 'pip', 'install', '--target', '/tmp/tmp1osq6e1x', '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Trainer-0.0+ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af-py3-none-any.whl']
running bdist_wheel
running build
running build_py
creating build
creating build/lib
copying taxi_transform.py -> build/lib
copying taxi_constants.py -> build/lib
copying taxi_trainer.py -> build/lib
running install
running install_lib
running install_egg_info
running egg_info
creating tfx_user_code_Trainer.egg-info
writing manifest file 'tfx_user_code_Trainer.egg-info/SOURCES.txt'
writing manifest file 'tfx_user_code_Trainer.egg-info/SOURCES.txt'
Copying tfx_user_code_Trainer.egg-info to /tmp/tmpbg9ly6tr/tfx_user_code_Trainer-0.0+ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af-py3.7.egg-info
running install_scripts
Processing /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Trainer-0.0+ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af-py3-none-any.whl
INFO:absl:Successfully installed '/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/_wheels/tfx_user_code_Trainer-0.0+ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af-py3-none-any.whl'.
INFO:absl:Training model.
INFO:absl:Feature company has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_census_tract has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_community_area has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_latitude has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_longitude has a shape . Setting to DenseTensor.
INFO:absl:Feature fare has a shape . Setting to DenseTensor.
INFO:absl:Feature payment_type has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_census_tract has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_community_area has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_latitude has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_longitude has a shape . Setting to DenseTensor.
INFO:absl:Feature tips has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_miles has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_seconds has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_day has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_hour has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_month has a shape . Setting to DenseTensor.
Installing collected packages: tfx-user-code-Trainer
Successfully installed tfx-user-code-Trainer-0.0+ace8eb563ff2ae66112acc05232b33344bcb925cdc0a0847df64c544323b99af
INFO:absl:Feature company has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_census_tract has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_community_area has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_latitude has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_longitude has a shape . Setting to DenseTensor.
INFO:absl:Feature fare has a shape . Setting to DenseTensor.
INFO:absl:Feature payment_type has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_census_tract has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_community_area has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_latitude has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_longitude has a shape . Setting to DenseTensor.
INFO:absl:Feature tips has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_miles has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_seconds has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_day has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_hour has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_month has a shape . Setting to DenseTensor.
INFO:absl:Feature company has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_census_tract has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_community_area has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_latitude has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_longitude has a shape . Setting to DenseTensor.
INFO:absl:Feature fare has a shape . Setting to DenseTensor.
INFO:absl:Feature payment_type has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_census_tract has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_community_area has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_latitude has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_longitude has a shape . Setting to DenseTensor.
INFO:absl:Feature tips has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_miles has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_seconds has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_day has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_hour has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_month has a shape . Setting to DenseTensor.
INFO:absl:Feature company has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_census_tract has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_community_area has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_latitude has a shape . Setting to DenseTensor.
INFO:absl:Feature dropoff_longitude has a shape . Setting to DenseTensor.
INFO:absl:Feature fare has a shape . Setting to DenseTensor.
INFO:absl:Feature payment_type has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_census_tract has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_community_area has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_latitude has a shape . Setting to DenseTensor.
INFO:absl:Feature pickup_longitude has a shape . Setting to DenseTensor.
INFO:absl:Feature tips has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_miles has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_seconds has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_day has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_hour has a shape . Setting to DenseTensor.
INFO:absl:Feature trip_start_month has a shape . Setting to DenseTensor.
/tmpfs/src/tf_docs_env/lib/python3.7/site-packages/keras/optimizer_v2/adam.py:105: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
super(Adam, self).__init__(name, **kwargs)
INFO:absl:Model: "model"
INFO:absl:__________________________________________________________________________________________________
INFO:absl: Layer (type) Output Shape Param # Connected to
INFO:absl:==================================================================================================
INFO:absl: company (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: dropoff_census_tract (InputLay [(None,)] 0 []
INFO:absl: er)
INFO:absl:
INFO:absl: dropoff_community_area (InputL [(None,)] 0 []
INFO:absl: ayer)
INFO:absl:
INFO:absl: dropoff_latitude (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: dropoff_longitude (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: fare (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: payment_type (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: pickup_census_tract (InputLaye [(None,)] 0 []
INFO:absl: r)
INFO:absl:
INFO:absl: pickup_community_area (InputLa [(None,)] 0 []
INFO:absl: yer)
INFO:absl:
INFO:absl: pickup_latitude (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: pickup_longitude (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: trip_miles (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: trip_seconds (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: trip_start_day (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: trip_start_hour (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: trip_start_month (InputLayer) [(None,)] 0 []
INFO:absl:
INFO:absl: dense_features (DenseFeatures) (None, 3) 0 ['company[0][0]',
INFO:absl: 'dropoff_census_tract[0][0]',
INFO:absl: 'dropoff_community_area[0][0]',
INFO:absl: 'dropoff_latitude[0][0]',
INFO:absl: 'dropoff_longitude[0][0]',
INFO:absl: 'fare[0][0]',
INFO:absl: 'payment_type[0][0]',
INFO:absl: 'pickup_census_tract[0][0]',
INFO:absl: 'pickup_community_area[0][0]',
INFO:absl: 'pickup_latitude[0][0]',
INFO:absl: 'pickup_longitude[0][0]',
INFO:absl: 'trip_miles[0][0]',
INFO:absl: 'trip_seconds[0][0]',
INFO:absl: 'trip_start_day[0][0]',
INFO:absl: 'trip_start_hour[0][0]',
INFO:absl: 'trip_start_month[0][0]']
INFO:absl:
INFO:absl: dense (Dense) (None, 100) 400 ['dense_features[0][0]']
INFO:absl:
INFO:absl: dense_1 (Dense) (None, 70) 7070 ['dense[0][0]']
INFO:absl:
INFO:absl: dense_2 (Dense) (None, 48) 3408 ['dense_1[0][0]']
INFO:absl:
INFO:absl: dense_3 (Dense) (None, 34) 1666 ['dense_2[0][0]']
INFO:absl:
INFO:absl: dense_features_1 (DenseFeature (None, 2127) 0 ['company[0][0]',
INFO:absl: s) 'dropoff_census_tract[0][0]',
INFO:absl: 'dropoff_community_area[0][0]',
INFO:absl: 'dropoff_latitude[0][0]',
INFO:absl: 'dropoff_longitude[0][0]',
INFO:absl: 'fare[0][0]',
INFO:absl: 'payment_type[0][0]',
INFO:absl: 'pickup_census_tract[0][0]',
INFO:absl: 'pickup_community_area[0][0]',
INFO:absl: 'pickup_latitude[0][0]',
INFO:absl: 'pickup_longitude[0][0]',
INFO:absl: 'trip_miles[0][0]',
INFO:absl: 'trip_seconds[0][0]',
INFO:absl: 'trip_start_day[0][0]',
INFO:absl: 'trip_start_hour[0][0]',
INFO:absl: 'trip_start_month[0][0]']
INFO:absl:
INFO:absl: concatenate (Concatenate) (None, 2161) 0 ['dense_3[0][0]',
INFO:absl: 'dense_features_1[0][0]']
INFO:absl:
INFO:absl: dense_4 (Dense) (None, 1) 2162 ['concatenate[0][0]']
INFO:absl:
INFO:absl:==================================================================================================
INFO:absl:Total params: 14,706
INFO:absl:Trainable params: 14,706
INFO:absl:Non-trainable params: 0
INFO:absl:__________________________________________________________________________________________________
10000/10000 [==============================] - 100s 10ms/step - loss: 0.2372 - binary_accuracy: 0.8605 - val_loss: 0.2222 - val_binary_accuracy: 0.8709
INFO:tensorflow:tensorflow_text is not available.
INFO:tensorflow:tensorflow_decision_forests is not available.
INFO:tensorflow:struct2tensor is not available.
WARNING:tensorflow:AutoGraph could not transform <bound method Socket.send of <zmq.Socket(zmq.PUSH) at 0x7f88b5e27910>> and will run it as-is.
Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.
Cause: module, class, method, function, traceback, frame, or code object was expected, got cython_function_or_method
To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert
WARNING: AutoGraph could not transform <bound method Socket.send of <zmq.Socket(zmq.PUSH) at 0x7f88b5e27910>> and will run it as-is.
Please report this to the TensorFlow team. When filing the bug, set the verbosity to 10 (on Linux, `export AUTOGRAPH_VERBOSITY=10`) and attach the full output.
Cause: module, class, method, function, traceback, frame, or code object was expected, got cython_function_or_method
To silence this warning, decorate the function with @tf.autograph.experimental.do_not_convert
INFO:absl:serve_transformed_features = {'pickup_latitude': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:9' shape=(None,) dtype=int64>, 'trip_start_hour': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:15' shape=(None,) dtype=int64>, 'fare': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:5' shape=(None,) dtype=float32>, 'trip_miles': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:12' shape=(None,) dtype=float32>, 'trip_start_day': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:14' shape=(None,) dtype=int64>, 'dropoff_latitude': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:3' shape=(None,) dtype=int64>, 'trip_start_month': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:16' shape=(None,) dtype=int64>, 'dropoff_community_area': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:2' shape=(None,) dtype=int64>, 'dropoff_longitude': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:4' shape=(None,) dtype=int64>, 'payment_type': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:6' shape=(None,) dtype=int64>, 'pickup_longitude': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:10' shape=(None,) dtype=int64>, 'pickup_community_area': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:8' shape=(None,) dtype=int64>, 'company': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:0' shape=(None,) dtype=int64>, 'pickup_census_tract': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:7' shape=(None,) dtype=int64>, 'dropoff_census_tract': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:1' shape=(None,) dtype=int64>, 'trip_seconds': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:13' shape=(None,) dtype=float32>}
INFO:absl:eval_transformed_features = {'pickup_latitude': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:9' shape=(None,) dtype=int64>, 'trip_start_hour': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:15' shape=(None,) dtype=int64>, 'fare': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:5' shape=(None,) dtype=float32>, 'trip_miles': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:12' shape=(None,) dtype=float32>, 'trip_start_day': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:14' shape=(None,) dtype=int64>, 'dropoff_latitude': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:3' shape=(None,) dtype=int64>, 'trip_start_month': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:16' shape=(None,) dtype=int64>, 'dropoff_community_area': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:2' shape=(None,) dtype=int64>, 'dropoff_longitude': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:4' shape=(None,) dtype=int64>, 'payment_type': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:6' shape=(None,) dtype=int64>, 'pickup_longitude': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:10' shape=(None,) dtype=int64>, 'pickup_community_area': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:8' shape=(None,) dtype=int64>, 'company': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:0' shape=(None,) dtype=int64>, 'pickup_census_tract': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:7' shape=(None,) dtype=int64>, 'tips': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:11' shape=(None,) dtype=int64>, 'dropoff_census_tract': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:1' shape=(None,) dtype=int64>, 'trip_seconds': <tf.Tensor 'transform_features_layer/StatefulPartitionedCall:13' shape=(None,) dtype=float32>}
INFO:tensorflow:Assets written to: /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Trainer/model/6/Format-Serving/assets
INFO:absl:Training complete. Model written to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Trainer/model/6/Format-Serving. ModelRun written to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Trainer/model_run/6
INFO:absl:Running publisher for Trainer
INFO:absl:MetadataStore with DB connection initialized
Analizuj szkolenie z TensorBoard
Rzuć okiem na artefakt trenera. Wskazuje na katalog zawierający podkatalogi modelu.
model_artifact_dir = trainer.outputs['model'].get()[0].uri
pp.pprint(os.listdir(model_artifact_dir))
model_dir = os.path.join(model_artifact_dir, 'Format-Serving')
pp.pprint(os.listdir(model_dir))
['Format-Serving'] ['variables', 'assets', 'keras_metadata.pb', 'saved_model.pb']
Opcjonalnie możemy podłączyć TensorBoard do Trainera, aby analizować krzywe treningowe naszego modelu.
model_run_artifact_dir = trainer.outputs['model_run'].get()[0].uri
%load_ext tensorboard
%tensorboard --logdir {model_run_artifact_dir}
Ewaluator
Evaluator składnik oblicza modelowych wskaźników osiąganych przez zestaw oceny. Używa TensorFlow analizy modelu biblioteki. Evaluator mogą także ewentualnie potwierdzić, że nowo przeszkoleni model jest lepszy niż w poprzednim modelu. Jest to przydatne w przypadku ustawienia potoku produkcyjnego, w którym można codziennie automatycznie trenować i weryfikować model. W tym notebooku, mamy tylko jeden model pociągu, więc Evaluator automatycznie oznaczyć model jako „dobre”.
Evaluator weźmie jako wejście dane z ExampleGen , przeszkolona model z Trainer i konfigurację krojenia. Konfiguracja wycinania umożliwia podzielenie metryk na wartości cech (np. jak Twój model radzi sobie z przejazdami taksówkami, które rozpoczynają się o godzinie 8:00 w porównaniu do 20:00?). Zobacz przykład takiej konfiguracji poniżej:
eval_config = tfma.EvalConfig(
model_specs=[
# This assumes a serving model with signature 'serving_default'. If
# using estimator based EvalSavedModel, add signature_name: 'eval' and
# remove the label_key.
tfma.ModelSpec(
signature_name='serving_default',
label_key='tips',
preprocessing_function_names=['transform_features'],
)
],
metrics_specs=[
tfma.MetricsSpec(
# The metrics added here are in addition to those saved with the
# model (assuming either a keras model or EvalSavedModel is used).
# Any metrics added into the saved model (for example using
# model.compile(..., metrics=[...]), etc) will be computed
# automatically.
# To add validation thresholds for metrics saved with the model,
# add them keyed by metric name to the thresholds map.
metrics=[
tfma.MetricConfig(class_name='ExampleCount'),
tfma.MetricConfig(class_name='BinaryAccuracy',
threshold=tfma.MetricThreshold(
value_threshold=tfma.GenericValueThreshold(
lower_bound={'value': 0.5}),
# Change threshold will be ignored if there is no
# baseline model resolved from MLMD (first run).
change_threshold=tfma.GenericChangeThreshold(
direction=tfma.MetricDirection.HIGHER_IS_BETTER,
absolute={'value': -1e-10})))
]
)
],
slicing_specs=[
# An empty slice spec means the overall slice, i.e. the whole dataset.
tfma.SlicingSpec(),
# Data can be sliced along a feature column. In this case, data is
# sliced along feature column trip_start_hour.
tfma.SlicingSpec(feature_keys=['trip_start_hour'])
])
Następnie dajemy tę konfigurację do Evaluator i uruchom go.
# Use TFMA to compute a evaluation statistics over features of a model and
# validate them against a baseline.
# The model resolver is only required if performing model validation in addition
# to evaluation. In this case we validate against the latest blessed model. If
# no model has been blessed before (as in this case) the evaluator will make our
# candidate the first blessed model.
model_resolver = tfx.dsl.Resolver(
strategy_class=tfx.dsl.experimental.LatestBlessedModelStrategy,
model=tfx.dsl.Channel(type=tfx.types.standard_artifacts.Model),
model_blessing=tfx.dsl.Channel(
type=tfx.types.standard_artifacts.ModelBlessing)).with_id(
'latest_blessed_model_resolver')
context.run(model_resolver)
evaluator = tfx.components.Evaluator(
examples=example_gen.outputs['examples'],
model=trainer.outputs['model'],
baseline_model=model_resolver.outputs['model'],
eval_config=eval_config)
context.run(evaluator)
INFO:absl:Running driver for latest_blessed_model_resolver
INFO:absl:MetadataStore with DB connection initialized
INFO:absl:Running publisher for latest_blessed_model_resolver
INFO:absl:MetadataStore with DB connection initialized
INFO:absl:Running driver for Evaluator
INFO:absl:MetadataStore with DB connection initialized
INFO:absl:Running executor for Evaluator
INFO:absl:Nonempty beam arg extra_packages already includes dependency
INFO:absl:udf_utils.get_fn {'eval_config': '{\n "metrics_specs": [\n {\n "metrics": [\n {\n "class_name": "ExampleCount"\n },\n {\n "class_name": "BinaryAccuracy",\n "threshold": {\n "change_threshold": {\n "absolute": -1e-10,\n "direction": "HIGHER_IS_BETTER"\n },\n "value_threshold": {\n "lower_bound": 0.5\n }\n }\n }\n ]\n }\n ],\n "model_specs": [\n {\n "label_key": "tips",\n "preprocessing_function_names": [\n "transform_features"\n ],\n "signature_name": "serving_default"\n }\n ],\n "slicing_specs": [\n {},\n {\n "feature_keys": [\n "trip_start_hour"\n ]\n }\n ]\n}', 'feature_slicing_spec': None, 'fairness_indicator_thresholds': 'null', 'example_splits': 'null', 'module_file': None, 'module_path': None} 'custom_eval_shared_model'
INFO:absl:Request was made to ignore the baseline ModelSpec and any change thresholds. This is likely because a baseline model was not provided: updated_config=
model_specs {
signature_name: "serving_default"
label_key: "tips"
preprocessing_function_names: "transform_features"
}
slicing_specs {
}
slicing_specs {
feature_keys: "trip_start_hour"
}
metrics_specs {
metrics {
class_name: "ExampleCount"
}
metrics {
class_name: "BinaryAccuracy"
threshold {
value_threshold {
lower_bound {
value: 0.5
}
}
}
}
}
INFO:absl:Using /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Trainer/model/6/Format-Serving as model.
WARNING:tensorflow:Inconsistent references when loading the checkpoint into this object graph. Either the Trackable object references in the Python program have changed in an incompatible way, or the checkpoint was generated in an incompatible program.
Two checkpoint references resolved to different objects (<keras.saving.saved_model.load.TensorFlowTransform>TransformFeaturesLayer object at 0x7f87bc0f5e50> and <keras.engine.input_layer.InputLayer object at 0x7f87bc0f5b50>).
INFO:absl:The 'example_splits' parameter is not set, using 'eval' split.
INFO:absl:Evaluating model.
INFO:absl:udf_utils.get_fn {'eval_config': '{\n "metrics_specs": [\n {\n "metrics": [\n {\n "class_name": "ExampleCount"\n },\n {\n "class_name": "BinaryAccuracy",\n "threshold": {\n "change_threshold": {\n "absolute": -1e-10,\n "direction": "HIGHER_IS_BETTER"\n },\n "value_threshold": {\n "lower_bound": 0.5\n }\n }\n }\n ]\n }\n ],\n "model_specs": [\n {\n "label_key": "tips",\n "preprocessing_function_names": [\n "transform_features"\n ],\n "signature_name": "serving_default"\n }\n ],\n "slicing_specs": [\n {},\n {\n "feature_keys": [\n "trip_start_hour"\n ]\n }\n ]\n}', 'feature_slicing_spec': None, 'fairness_indicator_thresholds': 'null', 'example_splits': 'null', 'module_file': None, 'module_path': None} 'custom_extractors'
INFO:absl:Request was made to ignore the baseline ModelSpec and any change thresholds. This is likely because a baseline model was not provided: updated_config=
model_specs {
signature_name: "serving_default"
label_key: "tips"
preprocessing_function_names: "transform_features"
}
slicing_specs {
}
slicing_specs {
feature_keys: "trip_start_hour"
}
metrics_specs {
metrics {
class_name: "ExampleCount"
}
metrics {
class_name: "BinaryAccuracy"
threshold {
value_threshold {
lower_bound {
value: 0.5
}
}
}
}
model_names: ""
}
INFO:absl:Request was made to ignore the baseline ModelSpec and any change thresholds. This is likely because a baseline model was not provided: updated_config=
model_specs {
signature_name: "serving_default"
label_key: "tips"
preprocessing_function_names: "transform_features"
}
slicing_specs {
}
slicing_specs {
feature_keys: "trip_start_hour"
}
metrics_specs {
metrics {
class_name: "ExampleCount"
}
metrics {
class_name: "BinaryAccuracy"
threshold {
value_threshold {
lower_bound {
value: 0.5
}
}
}
}
model_names: ""
}
INFO:absl:Request was made to ignore the baseline ModelSpec and any change thresholds. This is likely because a baseline model was not provided: updated_config=
model_specs {
signature_name: "serving_default"
label_key: "tips"
preprocessing_function_names: "transform_features"
}
slicing_specs {
}
slicing_specs {
feature_keys: "trip_start_hour"
}
metrics_specs {
metrics {
class_name: "ExampleCount"
}
metrics {
class_name: "BinaryAccuracy"
threshold {
value_threshold {
lower_bound {
value: 0.5
}
}
}
}
model_names: ""
}
WARNING:tensorflow:Inconsistent references when loading the checkpoint into this object graph. Either the Trackable object references in the Python program have changed in an incompatible way, or the checkpoint was generated in an incompatible program.
Two checkpoint references resolved to different objects (<keras.saving.saved_model.load.TensorFlowTransform>TransformFeaturesLayer object at 0x7f87b0102150> and <keras.engine.input_layer.InputLayer object at 0x7f875454e810>).
WARNING:root:Make sure that locally built Python SDK docker image has Python 3.7 interpreter.
WARNING:tensorflow:Inconsistent references when loading the checkpoint into this object graph. Either the Trackable object references in the Python program have changed in an incompatible way, or the checkpoint was generated in an incompatible program.
Two checkpoint references resolved to different objects (<keras.saving.saved_model.load.TensorFlowTransform>TransformFeaturesLayer object at 0x7f87b06c9d50> and <keras.engine.input_layer.InputLayer object at 0x7f87d4041290>).
WARNING:tensorflow:Inconsistent references when loading the checkpoint into this object graph. Either the Trackable object references in the Python program have changed in an incompatible way, or the checkpoint was generated in an incompatible program.
Two checkpoint references resolved to different objects (<keras.saving.saved_model.load.TensorFlowTransform>TransformFeaturesLayer object at 0x7f874c8d6a10> and <keras.engine.input_layer.InputLayer object at 0x7f874c8ac0d0>).
WARNING:tensorflow:Inconsistent references when loading the checkpoint into this object graph. Either the Trackable object references in the Python program have changed in an incompatible way, or the checkpoint was generated in an incompatible program.
Two checkpoint references resolved to different objects (<keras.saving.saved_model.load.TensorFlowTransform>TransformFeaturesLayer object at 0x7f830dcf9fd0> and <keras.engine.input_layer.InputLayer object at 0x7f830dd87110>).
WARNING:tensorflow:Inconsistent references when loading the checkpoint into this object graph. Either the Trackable object references in the Python program have changed in an incompatible way, or the checkpoint was generated in an incompatible program.
Two checkpoint references resolved to different objects (<keras.saving.saved_model.load.TensorFlowTransform>TransformFeaturesLayer object at 0x7f830dc8cad0> and <keras.engine.input_layer.InputLayer object at 0x7f830cf892d0>).
WARNING:tensorflow:Inconsistent references when loading the checkpoint into this object graph. Either the Trackable object references in the Python program have changed in an incompatible way, or the checkpoint was generated in an incompatible program.
Two checkpoint references resolved to different objects (<keras.saving.saved_model.load.TensorFlowTransform>TransformFeaturesLayer object at 0x7f87b041add0> and <keras.engine.input_layer.InputLayer object at 0x7f874d6b6d50>).
WARNING:tensorflow:Inconsistent references when loading the checkpoint into this object graph. Either the Trackable object references in the Python program have changed in an incompatible way, or the checkpoint was generated in an incompatible program.
Two checkpoint references resolved to different objects (<keras.saving.saved_model.load.TensorFlowTransform>TransformFeaturesLayer object at 0x7f830c42a5d0> and <keras.engine.input_layer.InputLayer object at 0x7f830c3037d0>).
INFO:absl:Evaluation complete. Results written to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Evaluator/evaluation/8.
INFO:absl:Checking validation results.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.7/site-packages/tensorflow_model_analysis/writers/metrics_plots_and_validations_writer.py:107: tf_record_iterator (from tensorflow.python.lib.io.tf_record) is deprecated and will be removed in a future version.
Instructions for updating:
Use eager execution and:
`tf.data.TFRecordDataset(path)`
INFO:absl:Blessing result True written to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Evaluator/blessing/8.
INFO:absl:Running publisher for Evaluator
INFO:absl:MetadataStore with DB connection initialized
Teraz zbadajmy artefakty wyjściowe Evaluator .
evaluator.outputs
{'evaluation': Channel(
type_name: ModelEvaluation
artifacts: [Artifact(artifact: id: 15
type_id: 29
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Evaluator/evaluation/8"
custom_properties {
key: "name"
value {
string_value: "evaluation"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Evaluator"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 29
name: "ModelEvaluation"
)]
additional_properties: {}
additional_custom_properties: {}
),
'blessing': Channel(
type_name: ModelBlessing
artifacts: [Artifact(artifact: id: 16
type_id: 30
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Evaluator/blessing/8"
custom_properties {
key: "blessed"
value {
int_value: 1
}
}
custom_properties {
key: "current_model"
value {
string_value: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Trainer/model/6"
}
}
custom_properties {
key: "current_model_id"
value {
int_value: 13
}
}
custom_properties {
key: "name"
value {
string_value: "blessing"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Evaluator"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 30
name: "ModelBlessing"
)]
additional_properties: {}
additional_custom_properties: {}
)}
Korzystanie z evaluation wyjście możemy pokazać domyślny wizualizację globalnych wskaźników na całym zestawie oceny.
context.show(evaluator.outputs['evaluation'])
Aby zobaczyć wizualizację dla wycinkowych metryk oceny, możemy bezpośrednio wywołać bibliotekę TensorFlow Model Analysis.
import tensorflow_model_analysis as tfma
# Get the TFMA output result path and load the result.
PATH_TO_RESULT = evaluator.outputs['evaluation'].get()[0].uri
tfma_result = tfma.load_eval_result(PATH_TO_RESULT)
# Show data sliced along feature column trip_start_hour.
tfma.view.render_slicing_metrics(
tfma_result, slicing_column='trip_start_hour')
SlicingMetricsViewer(config={'weightedExamplesColumn': 'example_count'}, data=[{'slice': 'trip_start_hour:19',…
Ta wizualizacja pokazuje te same dane, ale obliczany na każdej wartości charakterystycznej trip_start_hour zamiast na całego zestawu ewaluacyjnego.
Analiza modelu TensorFlow obsługuje wiele innych wizualizacji, takich jak wskaźniki rzetelności i wykreślanie szeregów czasowych wydajności modelu. Aby dowiedzieć się więcej, zobacz poradnik .
Ponieważ dodaliśmy progi do naszej konfiguracji, dostępne są również dane wyjściowe walidacji. Precence o blessing artefaktu oznacza, że nasz model przeszedł walidację. Ponieważ jest to pierwsza przeprowadzana walidacja, kandydat zostaje automatycznie pobłogosławiony.
blessing_uri = evaluator.outputs['blessing'].get()[0].uri
!ls -l {blessing_uri}
total 0 -rw-rw-r-- 1 kbuilder kbuilder 0 Dec 21 10:13 BLESSED
Teraz możesz również zweryfikować sukces, ładując rekord wyniku walidacji:
PATH_TO_RESULT = evaluator.outputs['evaluation'].get()[0].uri
print(tfma.load_validation_result(PATH_TO_RESULT))
validation_ok: true
validation_details {
slicing_details {
slicing_spec {
}
num_matching_slices: 25
}
}
Popychacz
Pusher składnikiem jest zwykle na końcu rurociągu TFX. Sprawdza, czy model przeszedł walidację, a jeśli tak, eksport modelu do _serving_model_dir .
pusher = tfx.components.Pusher(
model=trainer.outputs['model'],
model_blessing=evaluator.outputs['blessing'],
push_destination=tfx.proto.PushDestination(
filesystem=tfx.proto.PushDestination.Filesystem(
base_directory=_serving_model_dir)))
context.run(pusher)
INFO:absl:Running driver for Pusher INFO:absl:MetadataStore with DB connection initialized INFO:absl:Running executor for Pusher INFO:absl:Model version: 1640081600 INFO:absl:Model written to serving path /tmp/tmpkvhhk5j5/serving_model/taxi_simple/1640081600. INFO:absl:Model pushed to /tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Pusher/pushed_model/9. INFO:absl:Running publisher for Pusher INFO:absl:MetadataStore with DB connection initialized
Przeanalizujmy artefakty wyjściowe Pusher .
pusher.outputs
{'pushed_model': Channel(
type_name: PushedModel
artifacts: [Artifact(artifact: id: 17
type_id: 32
uri: "/tmp/tfx-interactive-2021-12-21T10_09_51.902969-bvucg0eq/Pusher/pushed_model/9"
custom_properties {
key: "name"
value {
string_value: "pushed_model"
}
}
custom_properties {
key: "producer_component"
value {
string_value: "Pusher"
}
}
custom_properties {
key: "pushed"
value {
int_value: 1
}
}
custom_properties {
key: "pushed_destination"
value {
string_value: "/tmp/tmpkvhhk5j5/serving_model/taxi_simple/1640081600"
}
}
custom_properties {
key: "pushed_version"
value {
string_value: "1640081600"
}
}
custom_properties {
key: "state"
value {
string_value: "published"
}
}
custom_properties {
key: "tfx_version"
value {
string_value: "1.5.0"
}
}
state: LIVE
, artifact_type: id: 32
name: "PushedModel"
)]
additional_properties: {}
additional_custom_properties: {}
)}
W szczególności Pusher wyeksportuje twój model w formacie SavedModel, który wygląda tak:
push_uri = pusher.outputs['pushed_model'].get()[0].uri
model = tf.saved_model.load(push_uri)
for item in model.signatures.items():
pp.pprint(item)
('serving_default',
<ConcreteFunction signature_wrapper(*, examples) at 0x7F82F31FDE50>)
('transform_features',
<ConcreteFunction signature_wrapper(*, examples) at 0x7F82F31AC410>)
Zakończyliśmy naszą wycieczkę po wbudowanych komponentach TFX!