Azure Databricksでpetastormやーる

投稿日 2022年4月27日
著者 azure-recipe-user
カテゴリー Azure
カテゴリー Databricks

はじめに

petastormやっていきまーす

開発環境

Spark から TensorFlow へのデータ変換を簡略化する

1．こちらのノートブックをやっていきます
https://docs.microsoft.com/ja-jp/azure/databricks/_static/notebooks/deep-learning/petastorm-spark-converter-tensorflow.html

2．ライブラリをインストール

%pip install petastorm
%pip install tensorflow
%pip install hyperopt
%pip install horovod
%pip install sparkdl
%pip install tensorframes

				
					
				1
2
3
4
5
6
7

						%pip install petastorm
%pip install tensorflow
%pip install hyperopt
%pip install horovod
%pip install sparkdl
%pip install tensorframes
 

					

			

3．ライブラリをインポート

from pyspark.sql.functions import col 
from petastorm.spark import SparkDatasetConverter, make_spark_converter
 
import io
import numpy as np
import tensorflow as tf
from PIL import Image
from petastorm import TransformSpec
from tensorflow import keras
from tensorflow.keras.applications.mobilenet_v2 import MobileNetV2, preprocess_input

from hyperopt import fmin, tpe, hp, SparkTrials, STATUS_OK 
import horovod.tensorflow.keras as hvd
from sparkdl import HorovodRunner

				
					
				1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

						from pyspark.sql.functions import col 
from petastorm.spark import SparkDatasetConverter, make_spark_converter
 
import io
import numpy as np
import tensorflow as tf
from PIL import Image
from petastorm import TransformSpec
from tensorflow import keras
from tensorflow.keras.applications.mobilenet_v2 import MobileNetV2, preprocess_input
 
from hyperopt import fmin, tpe, hp, SparkTrials, STATUS_OK 
import horovod.tensorflow.keras as hvd
from sparkdl import HorovodRunner
 

					

			

エラー出た

/databricks/python_shell/dbruntime/PythonPackageImportsInstrumentation/__init__.py:167: FutureWarning: pyarrow.LocalFileSystem is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  original_result = python_builtin_import(name, globals, locals, fromlist, level)
/local_disk0/.ephemeral_nfs/envs/pythonEnv-30dab267-e5ed-4c28-9b42-2c7adb85cdb4/lib/python3.8/site-packages/petastorm/spark/spark_dataset_converter.py:28: FutureWarning: pyarrow.LocalFileSystem is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  from pyarrow import LocalFileSystem
ImportError: cannot import name 'resnet50' from 'keras.applications' (/local_disk0/.ephemeral_nfs/envs/pythonEnv-30dab267-e5ed-4c28-9b42-2c7adb85cdb4/lib/python3.8/site-packages/keras/applications/__init__.py)

				
					
				1
2
3
4
5
6

						/databricks/python_shell/dbruntime/PythonPackageImportsInstrumentation/__init__.py:167: FutureWarning: pyarrow.LocalFileSystem is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  original_result = python_builtin_import(name, globals, locals, fromlist, level)
/local_disk0/.ephemeral_nfs/envs/pythonEnv-30dab267-e5ed-4c28-9b42-2c7adb85cdb4/lib/python3.8/site-packages/petastorm/spark/spark_dataset_converter.py:28: FutureWarning: pyarrow.LocalFileSystem is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  from pyarrow import LocalFileSystem
ImportError: cannot import name 'resnet50' from 'keras.applications' (/local_disk0/.ephemeral_nfs/envs/pythonEnv-30dab267-e5ed-4c28-9b42-2c7adb85cdb4/lib/python3.8/site-packages/keras/applications/__init__.py)
 

					

			

Spark から PyTorch ノートブックへのデータ変換を簡略化する

1．こちらのノートブックをやっていきます
https://docs.microsoft.com/ja-jp/azure/databricks/_static/notebooks/deep-learning/petastorm-spark-converter-pytorch.html

2．ライブラリをインストール

%pip install petastorm
%pip install torch==1.5.0
%pip install torchvision==0.6.0
%pip install hyperopt
%pip install horovod
%pip install sparkdl
%pip install tensorflow
%pip install tensorframes

				
					
				1
2
3
4
5
6
7
8
9

						%pip install petastorm
%pip install torch==1.5.0
%pip install torchvision==0.6.0
%pip install hyperopt
%pip install horovod
%pip install sparkdl
%pip install tensorflow
%pip install tensorframes
 

					

			

3．ライブラリをインポート

from pyspark.sql.functions import col
 
from petastorm.spark import SparkDatasetConverter, make_spark_converter
 
import io
import numpy as np
import torch
import torchvision
from PIL import Image
from functools import partial 
from petastorm import TransformSpec
from torchvision import transforms
 
from hyperopt import fmin, tpe, hp, SparkTrials, STATUS_OK
 
import horovod.torch as hvd
from sparkdl import HorovodRunner

				
					
				1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

						from pyspark.sql.functions import col
 
from petastorm.spark import SparkDatasetConverter, make_spark_converter
 
import io
import numpy as np
import torch
import torchvision
from PIL import Image
from functools import partial 
from petastorm import TransformSpec
from torchvision import transforms
 
from hyperopt import fmin, tpe, hp, SparkTrials, STATUS_OK
 
import horovod.torch as hvd
from sparkdl import HorovodRunner
 

					

			

エラー出た

/databricks/python_shell/dbruntime/PythonPackageImportsInstrumentation/__init__.py:167: FutureWarning: pyarrow.LocalFileSystem is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  original_result = python_builtin_import(name, globals, locals, fromlist, level)
/local_disk0/.ephemeral_nfs/envs/pythonEnv-ba09db16-2f6f-4866-949c-4d5aa9fbed19/lib/python3.8/site-packages/petastorm/spark/spark_dataset_converter.py:28: FutureWarning: pyarrow.LocalFileSystem is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  from pyarrow import LocalFileSystem
Extension horovod.torch has not been built: /local_disk0/.ephemeral_nfs/envs/pythonEnv-ba09db16-2f6f-4866-949c-4d5aa9fbed19/lib/python3.8/site-packages/horovod/torch/mpi_lib/_mpi_lib.cpython-38-x86_64-linux-gnu.so not found
If this is not expected, reinstall Horovod with HOROVOD_WITH_PYTORCH=1 to debug the build error.
Warning! MPI libs are missing, but python applications are still available.
ImportError: cannot import name 'resnet50' from 'keras.applications' (/local_disk0/.ephemeral_nfs/envs/pythonEnv-ba09db16-2f6f-4866-949c-4d5aa9fbed19/lib/python3.8/site-packages/keras/applications/__init__.py)

				
					
				1
2
3
4
5
6
7
8
9

						/databricks/python_shell/dbruntime/PythonPackageImportsInstrumentation/__init__.py:167: FutureWarning: pyarrow.LocalFileSystem is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  original_result = python_builtin_import(name, globals, locals, fromlist, level)
/local_disk0/.ephemeral_nfs/envs/pythonEnv-ba09db16-2f6f-4866-949c-4d5aa9fbed19/lib/python3.8/site-packages/petastorm/spark/spark_dataset_converter.py:28: FutureWarning: pyarrow.LocalFileSystem is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  from pyarrow import LocalFileSystem
Extension horovod.torch has not been built: /local_disk0/.ephemeral_nfs/envs/pythonEnv-ba09db16-2f6f-4866-949c-4d5aa9fbed19/lib/python3.8/site-packages/horovod/torch/mpi_lib/_mpi_lib.cpython-38-x86_64-linux-gnu.so not found
If this is not expected, reinstall Horovod with HOROVOD_WITH_PYTORCH=1 to debug the build error.
Warning! MPI libs are missing, but python applications are still available.
ImportError: cannot import name 'resnet50' from 'keras.applications' (/local_disk0/.ephemeral_nfs/envs/pythonEnv-ba09db16-2f6f-4866-949c-4d5aa9fbed19/lib/python3.8/site-packages/keras/applications/__init__.py)
 

					

			

Spark と Petastorm を使用してディープラーニングノートブック用のデータを準備する

1．こちらのノートブックをやっていきます

petastorm – Databricks

2．ライブラリをインストール

%pip install tensorflow
%pip install petastorm

				1
2
3

						%pip install tensorflow
%pip install petastorm

3．ライブラリをインポート

import os
import subprocess
import uuid

				1
2
3
4

						import os
import subprocess
import uuid

3．ディレクトリを作成

# Set a unique working directory for this notebook.
work_dir = os.path.join("/ml/tmp/petastorm", str(uuid.uuid4()))
dbutils.fs.mkdirs(work_dir)
 
def get_local_path(dbfs_path):
  return os.path.join("/dbfs", dbfs_path.lstrip("/"))

				
					
				1
2
3
4
5
6
7

						# Set a unique working directory for this notebook.
work_dir = os.path.join("/ml/tmp/petastorm", str(uuid.uuid4()))
dbutils.fs.mkdirs(work_dir)
 
def get_local_path(dbfs_path):
  return os.path.join("/dbfs", dbfs_path.lstrip("/"))
 

					

			

4．データセット（mnist）をダウンロード

data_url = "https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/mnist.bz2"
libsvm_path = os.path.join(work_dir, "mnist.bz2")
subprocess.check_output(["wget", data_url, "-O", get_local_path(libsvm_path)])

				1
2
3
4

						data_url = "https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/mnist.bz2"
libsvm_path = os.path.join(work_dir, "mnist.bz2")
subprocess.check_output(["wget", data_url, "-O", get_local_path(libsvm_path)])

Out[13]: b''

				1
2

						Out[13]: b''

5．データセットの読み込み

df = spark.read.format("libsvm") \
  .option("numFeatures", "784") \
  .load(libsvm_path)

				1
2
3
4

						df = spark.read.format("libsvm") \
  .option("numFeatures", "784") \
  .load(libsvm_path)

6．データを配列に格納

%scala
 
import org.apache.spark.ml.linalg.Vector
 
val toArray = udf { v: Vector =&gt; v.toArray }
spark.udf.register("toArray", toArray)

				1
2
3
4
5
6
7

						%scala
 
import org.apache.spark.ml.linalg.Vector
 
val toArray = udf { v: Vector =&gt; v.toArray }
spark.udf.register("toArray", toArray)

import org.apache.spark.ml.linalg.Vector
toArray: org.apache.spark.sql.expressions.UserDefinedFunction = SparkUserDefinedFunction($Lambda$10093/495988232@1dd62a2c,ArrayType(DoubleType,false),List(Some(class[value[0]: vector])),Some(class[value[0]: array&lt;double&gt;]),None,true,true)
res1: org.apache.spark.sql.expressions.UserDefinedFunction = SparkUserDefinedFunction($Lambda$10093/495988232@1dd62a2c,ArrayType(DoubleType,false),List(Some(class[value[0]: vector])),Some(class[value[0]: array&lt;double&gt;]),Some(toArray),true,true)

				1
2
3
4

						import org.apache.spark.ml.linalg.Vector
toArray: org.apache.spark.sql.expressions.UserDefinedFunction = SparkUserDefinedFunction($Lambda$10093/495988232@1dd62a2c,ArrayType(DoubleType,false),List(Some(class[value[0]: vector])),Some(class[value[0]: array&lt;double&gt;]),None,true,true)
res1: org.apache.spark.sql.expressions.UserDefinedFunction = SparkUserDefinedFunction($Lambda$10093/495988232@1dd62a2c,ArrayType(DoubleType,false),List(Some(class[value[0]: vector])),Some(class[value[0]: array&lt;double&gt;]),Some(toArray),true,true)

7．Parquet形式へ変換

# Convert sparse vectors to dense arrays and write the data as Parquet.
# Petastorm will sample Parquet row groups into batches.
# Batch size is important for the utilization of both I/O and compute.
# You can use parquet.block.size to control the size.
parquet_path = os.path.join(work_dir, "parquet")
df.selectExpr("toArray(features) AS features", "int(label) AS label") \
  .repartition(10) \
  .write.mode("overwrite") \
  .option("parquet.block.size", 1024 * 1024) \
  .parquet(parquet_path)

				
					
				1
2
3
4
5
6
7
8
9
10
11

						# Convert sparse vectors to dense arrays and write the data as Parquet.
# Petastorm will sample Parquet row groups into batches.
# Batch size is important for the utilization of both I/O and compute.
# You can use parquet.block.size to control the size.
parquet_path = os.path.join(work_dir, "parquet")
df.selectExpr("toArray(features) AS features", "int(label) AS label") \
  .repartition(10) \
  .write.mode("overwrite") \
  .option("parquet.block.size", 1024 * 1024) \
  .parquet(parquet_path)
 

					

			

8．ライブラリをインポート

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import models, layers
 
from petastorm import make_batch_reader
from petastorm.tf_utils import make_petastorm_dataset

				
					
				1
2
3
4
5
6
7

						import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import models, layers
 
from petastorm import make_batch_reader
from petastorm.tf_utils import make_petastorm_dataset
 

					

			

9．モデルの定義

def get_model():
  model = models.Sequential()
  model.add(layers.Conv2D(32, kernel_size=(3, 3),
                          activation='relu',
                          input_shape=(28, 28, 1)))
  model.add(layers.Conv2D(64, (3, 3), activation='relu'))
  model.add(layers.MaxPooling2D(pool_size=(2, 2)))
  model.add(layers.Dropout(0.25))
  model.add(layers.Flatten())
  model.add(layers.Dense(128, activation='relu'))
  model.add(layers.Dropout(0.5))
  model.add(layers.Dense(10, activation='softmax'))
  return model

				
					
				1
2
3
4
5
6
7
8
9
10
11
12
13
14

						def get_model():
  model = models.Sequential()
  model.add(layers.Conv2D(32, kernel_size=(3, 3),
                          activation='relu',
                          input_shape=(28, 28, 1)))
  model.add(layers.Conv2D(64, (3, 3), activation='relu'))
  model.add(layers.MaxPooling2D(pool_size=(2, 2)))
  model.add(layers.Dropout(0.25))
  model.add(layers.Flatten())
  model.add(layers.Dense(128, activation='relu'))
  model.add(layers.Dropout(0.5))
  model.add(layers.Dense(10, activation='softmax'))
  return model
 

					

			

10．parquet形式のデータのパスを定義

import os
import pyarrow.parquet as pq

underscore_files = [f for f in os.listdir(get_local_path(parquet_path)) if f.startswith("_")]
pq.EXCLUDED_PARQUET_PATHS.update(underscore_files)

				1
2
3
4
5
6

						import os
import pyarrow.parquet as pq
 
underscore_files = [f for f in os.listdir(get_local_path(parquet_path)) if f.startswith("_")]
pq.EXCLUDED_PARQUET_PATHS.update(underscore_files)

11．Parquet形式のデータをmake_batch_readerで読み取り、学習

# We use make_batch_reader to load Parquet row groups into batches.
# HINT: Use cur_shard and shard_count params to shard data in distributed training.
petastorm_dataset_url = "file://" + get_local_path(parquet_path)
with make_batch_reader(petastorm_dataset_url, num_epochs=100) as reader:
  dataset = make_petastorm_dataset(reader) \
    .map(lambda x: (tf.reshape(x.features, [-1, 28, 28, 1]), tf.one_hot(x.label, 10)))
  model = get_model()
  optimizer = keras.optimizers.Adadelta()
  model.compile(optimizer=optimizer,
                loss='categorical_crossentropy',
                metrics=['accuracy'])
  model.fit(dataset, steps_per_epoch=10, epochs=10)

				
					
				1
2
3
4
5
6
7
8
9
10
11
12
13

						# We use make_batch_reader to load Parquet row groups into batches.
# HINT: Use cur_shard and shard_count params to shard data in distributed training.
petastorm_dataset_url = "file://" + get_local_path(parquet_path)
with make_batch_reader(petastorm_dataset_url, num_epochs=100) as reader:
  dataset = make_petastorm_dataset(reader) \
    .map(lambda x: (tf.reshape(x.features, [-1, 28, 28, 1]), tf.one_hot(x.label, 10)))
  model = get_model()
  optimizer = keras.optimizers.Adadelta()
  model.compile(optimizer=optimizer,
                loss='categorical_crossentropy',
                metrics=['accuracy'])
  model.fit(dataset, steps_per_epoch=10, epochs=10)
 

					

			

/local_disk0/.ephemeral_nfs/envs/pythonEnv-480ab404-1e1a-48b7-b616-46583213ddef/lib/python3.8/site-packages/petastorm/fs_utils.py:88: FutureWarning: pyarrow.localfs is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  self._filesystem = pyarrow.localfs
Epoch 1/10
10/10 [==============================] - 3s 159ms/step - loss: 42.4297 - accuracy: 0.0952
Epoch 2/10
...

				
					
				1
2
3
4
5
6
7

						/local_disk0/.ephemeral_nfs/envs/pythonEnv-480ab404-1e1a-48b7-b616-46583213ddef/lib/python3.8/site-packages/petastorm/fs_utils.py:88: FutureWarning: pyarrow.localfs is deprecated as of 2.0.0, please use pyarrow.fs.LocalFileSystem instead.
  self._filesystem = pyarrow.localfs
Epoch 1/10
10/10 [==============================] - 3s 159ms/step - loss: 42.4297 - accuracy: 0.0952
Epoch 2/10
...
 

					

			

12．ディレクトリの削除

# Clean up the working directory.
dbutils.fs.rm(work_dir, recurse=True)

				1
2
3

						# Clean up the working directory.
dbutils.fs.rm(work_dir, recurse=True)

13．ディレクトリが残っていた場合の確認

%fs ls /ml/tmp/petastorm

				1
2

						%fs ls /ml/tmp/petastorm

path,name,size,modificationTime
dbfs:/ml/tmp/petastorm/7f3d8aa0-0ecf-4b75-b649-f13a8259f2bd/,7f3d8aa0-0ecf-4b75-b649-f13a8259f2bd/,0,1650708482000

				1
2
3

						path,name,size,modificationTime
dbfs:/ml/tmp/petastorm/7f3d8aa0-0ecf-4b75-b649-f13a8259f2bd/,7f3d8aa0-0ecf-4b75-b649-f13a8259f2bd/,0,1650708482000

14．uuidを指定して削除

dbutils.fs.rm("/ml/tmp/petastorm/7f3d8aa0-0ecf-4b75-b649-f13a8259f2bd/", recurse=True)

				1
2

						dbutils.fs.rm("/ml/tmp/petastorm/7f3d8aa0-0ecf-4b75-b649-f13a8259f2bd/", recurse=True)

参考文献

Petastorm を使用してデータを読み込む – Azure Databricks

Azure Databricksの導入ならナレコムにおまかせください。

導入から活用方法までサポートします。お気軽にご相談ください。

Azure Databricksソリューションページはこちら

この記事を書いた人

azure-recipe-user

記事一覧

Azure Databricksでpetastormやーる

はじめに

開発環境

Spark から TensorFlow へのデータ変換を簡略化する

Spark から PyTorch ノートブックへのデータ変換を簡略化する

Spark と Petastorm を使用してディープラーニングノートブック用のデータを準備する

参考文献

Azure Databricksの導入ならナレコムにおまかせください。

この記事を書いた人

azure-recipe-user

５分で学ぶAzure LogAnalytics【中級者向け】

Azure Databricks: 3-2. DBFSにAzure Data Lake Storage Gen2をマウント

Databricks Named a Leader in Gartner Magic Quadrant for Data Science and Machine Learning Platforms　翻訳してみた

APIキーも安全に！Azure DatabricksとAzure OpenAI Serviceの連携術

はじめに

開発環境

Spark から TensorFlow へのデータ変換を簡略化する

Spark から PyTorch ノートブックへのデータ変換を簡略化する

Spark と Petastorm を使用してディープ ラーニング ノートブック用のデータを準備する

参考文献

Azure Databricksの導入ならナレコムにおまかせください。

この記事を書いた人

azure-recipe-user

Spark と Petastorm を使用してディープラーニングノートブック用のデータを準備する