Databricks Runtime 7.0 ML (non pris en charge)

Databricks a publié cette image en juin 2020.

Databricks Runtime 7.0 for Machine Learning fournit un environnement prêt à l'emploi pour l'apprentissage automatique et la science des données basé sur Databricks Runtime 7.0 (non pris en charge). Databricks Runtime ML contient de nombreuses bibliothèques populaires de Machine Learning, notamment TensorFlow, PyTorch et XGBoost. Il prend également en charge la formation de Deep Learning distribué avec Horovod.

Pour plus d’informations, notamment les instructions relatives à la création d’un cluster Databricks Runtime ML, consultez IA et Machine Learning sur Databricks.

Nouvelles fonctionnalités et modifications majeures

Databricks Runtime 7.0 ML est basé sur Databricks Runtime 7.0. Pour plus d'informations sur les nouveautés de Databricks Runtime 7.0, y compris Apache Spark MLlib et SparkR, consultez les notes de publication de Databricks Runtime 7.0 (non pris en charge).

Planification compatible GPU

Databricks Runtime 7.0 ML prend en charge la planification compatible GPU à partir d’Apache Spark 3.0. Azure Databricks la configure automatiquement. Consultez Planification de GPU.

Modifications majeures apportées à l’environnement ML Python

Cette section décrit les modifications majeures apportées à l’environnement ML Python préinstallé par rapport à Databricks Runtime 6.6 ML (sans support). Vous devez également passer en revue les modifications majeures apportées à l’environnement Python de base dans Databricks Runtime 7.0 (sans support). Pour obtenir la liste complète des packages Python installés et leurs versions, consultez Bibliothèques Python.

Mise à niveau des packages Python

  • tensorflow 1.15.0 -> 2.2.0
  • tensorboard 1.15.0 -> 2.2.2
  • pytorch 1.4.0 -> 1.5.0
  • xgboost 0.90 -> 1.1.1
  • sparkdl 1.6.0-db1 -> 2.1.0-db1
  • hyperopt 0.2.2.db1 -> 0.2.4.db1

Ajout de packages Python

  • lightgbm : 2.3.0
  • nltk : 3.4.5
  • petastorm : 0.9.2
  • plotly : 4.5.2

Packages Python supprimés

  • argparse
  • boto (utiliser boto3 à la place)
  • colorama
  • déconseillé
  • et-xmlfile
  • fusepy
  • html5lib
  • jdcal
  • keras (utiliser tensorflow.keras à la place)
  • keras-applications (utiliser tensorflow.keras.applications à la place)
  • llvmlite
  • lxml
  • nose
  • nose-exclude
  • numba
  • openpyxl
  • pathlib2
  • ply
  • pymongo
  • singledispatch
  • tensorboardX (utiliser torch.utils.tensorboard à la place)
  • virtualenv
  • webencodings

Modifications majeures apportées à l’environnement R ML

Databricks Runtime 7.0 ML contient une version non modifiée de RStudio Server open source v1.2.5033 pour laquelle le code source se trouve dans GitHub. Découvrez-en plus sur RStudio Server sur Azure Databricks.

Modifications apportées aux packages ML Spark, et aux bibliothèques Java et Scala

Les packages suivants sont mis à niveau. Certains sont mis à niveau vers des versions de SNAPSHOT qui sont compatibles avec Apache Spark 3.0 :

  • graphframes : 0.7.0-db1-spark2.4 -> 0.8.0-db2-spark3.0
  • spark-tensorflow-connector : 1.15.0 (Scala 2.11) -> 1.15.0 (Scala 2.12)
  • xgboost4j et xgboost4j-spark : 0.90 -> 1.0.0
  • mleap-databricks-runtime : 0.17.0-4882dc3 (SNAPSHOT)

Les packages suivants sont supprimés :

  • TensorFlow (Java)
  • TensorFrames
  • Pipelines de Deep Learning pour Apache Spark (HorovodRunner est disponible dans Python)

Ajout de commandes conda et pip pour prendre en charge les bibliothèques Python délimitées aux notebooks (préversion publique)

À compter de Databricks Runtime 7.0 ML, vous pouvez utiliser les commandes %pip et %conda pour gérer les bibliothèques Python installées dans une session de notebook. Vous pouvez également utiliser ces commandes pour créer un environnement personnalisé pour un notebook et pour reproduire cet environnement entre plusieurs notebooks. Pour activer cette fonctionnalité, dans les paramètres de cluster, définissez la configuration Sparkspark.databricks.conda.condaMagic.enabled true. Pour plus d’informations, consultez Bibliothèques Python délimitées aux notebooks.

Dépréciations et fonctionnalités non prises en charge

Databricks Runtime 7.0 ML ne prend pas en charge le contrôle d’accès aux tables. Si vous avez besoin d’un contrôle d’accès aux tables, nous vous recommandons d’utiliser Databricks Runtime 7.0.

Problèmes connus

  • Si vous journaliser un modèle MLlib au format mleap, lorsque l’argument sample_input est transmis à mlflow.spark.log_model, celui-ci échoue avec un AttributeError. Ce problème est dû à une modification de l’API vers mleap. Pour contourner ce problème, effectuez une mise à niveau vers MLflow 1.9.0. Vous pouvez installer MLflow 1.9.0 à l’aide de bibliothèques Python délimitées aux notebooks.

Environnement du système

L’environnement système de Databricks Runtime 7.0 ML diffère de Databricks Runtime 7.0 comme suit :

Bibliothèques

Les sections suivantes listent les bibliothèques incluses dans Databricks Runtime 7.0 ML qui diffèrent de celles incluses dans Databricks Runtime 7.0.

Dans cette section :

Bibliothèques de niveau supérieur

Databricks Runtime 7.0 ML comprend les bibliothèques de niveau supérieur suivantes :

Bibliothèques Python

Databricks Runtime 7.0 ML utilise Conda pour la gestion des packages Python et comprend de nombreux packages ML populaires. La section suivante décrit l’environnement Conda pour Databricks Runtime 7.0 ML.

Python sur les clusters UC

name: databricks-ml
channels:
  - pytorch
  - defaults
dependencies:
  - _libgcc_mutex=0.1=main
  - absl-py=0.9.0=py37_0
  - asn1crypto=1.3.0=py37_0
  - astor=0.8.0=py37_0
  - backcall=0.1.0=py37_0
  - backports=1.0=py_2
  - bcrypt=3.1.7=py37h7b6447c_1
  - blas=1.0=mkl
  - blinker=1.4=py37_0
  - boto3=1.12.0=py_0
  - botocore=1.15.0=py_0
  - c-ares=1.15.0=h7b6447c_1001
  - ca-certificates=2020.1.1=0
  - cachetools=4.1.0=py_1
  - certifi=2020.4.5.1=py37_0
  - cffi=1.14.0=py37h2e261b9_0
  - chardet=3.0.4=py37_1003
  - click=7.0=py37_0
  - cloudpickle=1.3.0=py_0
  - configparser=3.7.4=py37_0
  - cpuonly=1.0=0
  - cryptography=2.8=py37h1ba5d50_0
  - cycler=0.10.0=py37_0
  - cython=0.29.15=py37he6710b0_0
  - decorator=4.4.1=py_0
  - dill=0.3.1.1=py37_1
  - docutils=0.15.2=py37_0
  - entrypoints=0.3=py37_0
  - flask=1.1.1=py_1
  - freetype=2.9.1=h8a8886c_1
  - future=0.18.2=py37_1
  - gast=0.3.3=py_0
  - gitdb2=2.0.6=py_0
  - gitpython=3.0.5=py_0
  - google-auth=1.11.2=py_0
  - google-auth-oauthlib=0.4.1=py_2
  - google-pasta=0.2.0=py_0
  - grpcio=1.27.2=py37hf8bcb03_0
  - gunicorn=20.0.4=py37_0
  - h5py=2.10.0=py37h7918eee_0
  - hdf5=1.10.4=hb1b8bf9_0
  - icu=58.2=he6710b0_3
  - idna=2.8=py37_0
  - intel-openmp=2020.0=166
  - ipykernel=5.1.4=py37h39e3cac_0
  - ipython=7.12.0=py37h5ca1d4c_0
  - ipython_genutils=0.2.0=py37_0
  - itsdangerous=1.1.0=py37_0
  - jedi=0.14.1=py37_0
  - jinja2=2.11.1=py_0
  - jmespath=0.9.4=py_0
  - joblib=0.14.1=py_0
  - jpeg=9b=h024ee3a_2
  - jupyter_client=5.3.4=py37_0
  - jupyter_core=4.6.1=py37_0
  - kiwisolver=1.1.0=py37he6710b0_0
  - krb5=1.16.4=h173b8e3_0
  - ld_impl_linux-64=2.33.1=h53a641e_7
  - libedit=3.1.20181209=hc058e9b_0
  - libffi=3.2.1=hd88cf55_4
  - libgcc-ng=9.1.0=hdf63c60_0
  - libgfortran-ng=7.3.0=hdf63c60_0
  - libpng=1.6.37=hbc83047_0
  - libpq=11.2=h20c2e04_0
  - libprotobuf=3.11.4=hd408876_0
  - libsodium=1.0.16=h1bed415_0
  - libstdcxx-ng=9.1.0=hdf63c60_0
  - libtiff=4.1.0=h2733197_0
  - lightgbm=2.3.0=py37he6710b0_0
  - lz4-c=1.8.1.2=h14c3975_0
  - mako=1.1.2=py_0
  - markdown=3.1.1=py37_0
  - markupsafe=1.1.1=py37h7b6447c_0
  - matplotlib-base=3.1.3=py37hef1b27d_0
  - mkl=2020.0=166
  - mkl-service=2.3.0=py37he904b0f_0
  - mkl_fft=1.0.15=py37ha843d7b_0
  - mkl_random=1.1.0=py37hd6b4f25_0
  - ncurses=6.2=he6710b0_1
  - networkx=2.4=py_0
  - ninja=1.9.0=py37hfd86e86_0
  - nltk=3.4.5=py37_0
  - numpy=1.18.1=py37h4f9e942_0
  - numpy-base=1.18.1=py37hde5b4d6_1
  - oauthlib=3.1.0=py_0
  - olefile=0.46=py37_0
  - openssl=1.1.1g=h7b6447c_0
  - packaging=20.1=py_0
  - pandas=1.0.1=py37h0573a6f_0
  - paramiko=2.7.1=py_0
  - parso=0.5.2=py_0
  - patsy=0.5.1=py37_0
  - pexpect=4.8.0=py37_0
  - pickleshare=0.7.5=py37_0
  - pillow=7.0.0=py37hb39fc2d_0
  - pip=20.0.2=py37_3
  - plotly=4.5.2=py_0
  - prompt_toolkit=3.0.3=py_0
  - protobuf=3.11.4=py37he6710b0_0
  - psutil=5.6.7=py37h7b6447c_0
  - psycopg2=2.8.4=py37h1ba5d50_0
  - ptyprocess=0.6.0=py37_0
  - pyasn1=0.4.8=py_0
  - pyasn1-modules=0.2.7=py_0
  - pycparser=2.19=py37_0
  - pygments=2.5.2=py_0
  - pyjwt=1.7.1=py37_0
  - pynacl=1.3.0=py37h7b6447c_0
  - pyodbc=4.0.30=py37he6710b0_0
  - pyopenssl=19.1.0=py37_0
  - pyparsing=2.4.6=py_0
  - pysocks=1.7.1=py37_0
  - python=3.7.6=h0371630_2
  - python-dateutil=2.8.1=py_0
  - python-editor=1.0.4=py_0
  - pytorch=1.5.0=py3.7_cpu_0
  - pytz=2019.3=py_0
  - pyzmq=18.1.1=py37he6710b0_0
  - readline=7.0=h7b6447c_5
  - requests=2.22.0=py37_1
  - requests-oauthlib=1.3.0=py_0
  - retrying=1.3.3=py37_2
  - rsa=4.0=py_0
  - s3transfer=0.3.3=py37_0
  - scikit-learn=0.22.1=py37hd81dba3_0
  - scipy=1.4.1=py37h0b6359f_0
  - setuptools=45.2.0=py37_0
  - simplejson=3.17.0=py37h7b6447c_0
  - six=1.14.0=py37_0
  - smmap2=2.0.5=py37_0
  - sqlite=3.31.1=h62c20be_1
  - sqlparse=0.3.0=py_0
  - statsmodels=0.11.0=py37h7b6447c_0
  - tabulate=0.8.3=py37_0
  - tk=8.6.8=hbc83047_0
  - torchvision=0.6.0=py37_cpu
  - tornado=6.0.3=py37h7b6447c_3
  - tqdm=4.42.1=py_0
  - traitlets=4.3.3=py37_0
  - unixodbc=2.3.7=h14c3975_0
  - urllib3=1.25.8=py37_0
  - wcwidth=0.1.8=py_0
  - websocket-client=0.56.0=py37_0
  - werkzeug=1.0.0=py_0
  - wheel=0.34.2=py37_0
  - wrapt=1.11.2=py37h7b6447c_0
  - xz=5.2.4=h14c3975_4
  - zeromq=4.3.1=he6710b0_3
  - zlib=1.2.11=h7b6447c_3
  - zstd=1.3.7=h0b5b093_0
  - pip:
    - astunparse==1.6.3
    - databricks-cli==0.11.0
    - diskcache==4.1.0
    - docker==4.2.1
    - gorilla==0.3.0
    - horovod==0.19.1
    - hyperopt==0.2.4.db1
    - keras-preprocessing==1.1.2
    - mleap==0.16.0
    - mlflow==1.8.0
    - opt-einsum==3.2.1
    - petastorm==0.9.2
    - pyarrow==0.15.1
    - pyyaml==5.3.1
    - querystring-parser==1.2.4
    - seaborn==0.10.0
    - sparkdl==2.1.0-db1
    - tensorboard==2.2.2
    - tensorboard-plugin-wit==1.6.0.post3
    - tensorflow-cpu==2.2.0
    - tensorflow-estimator==2.2.0
    - termcolor==1.1.0
    - xgboost==1.1.1
prefix: /databricks/conda/envs/databricks-ml

Python sur les clusters GPU

name: databricks-ml-gpu
channels:
  - pytorch
  - defaults
dependencies:
  - _libgcc_mutex=0.1=main
  - absl-py=0.9.0=py37_0
  - asn1crypto=1.3.0=py37_0
  - astor=0.8.0=py37_0
  - backcall=0.1.0=py37_0
  - backports=1.0=py_2
  - bcrypt=3.1.7=py37h7b6447c_1
  - blas=1.0=mkl
  - blinker=1.4=py37_0
  - boto3=1.12.0=py_0
  - botocore=1.15.0=py_0
  - c-ares=1.15.0=h7b6447c_1001
  - ca-certificates=2020.1.1=0
  - cachetools=4.1.0=py_1
  - certifi=2020.4.5.2=py37_0
  - cffi=1.14.0=py37h2e261b9_0
  - chardet=3.0.4=py37_1003
  - click=7.0=py37_0
  - cloudpickle=1.3.0=py_0
  - configparser=3.7.4=py37_0
  - cryptography=2.8=py37h1ba5d50_0
  - cudatoolkit=10.1.243=h6bb024c_0
  - cycler=0.10.0=py37_0
  - cython=0.29.15=py37he6710b0_0
  - decorator=4.4.1=py_0
  - dill=0.3.1.1=py37_1
  - docutils=0.15.2=py37_0
  - entrypoints=0.3=py37_0
  - flask=1.1.1=py_1
  - freetype=2.9.1=h8a8886c_1
  - future=0.18.2=py37_1
  - gast=0.3.3=py_0
  - gitdb2=2.0.6=py_0
  - gitpython=3.0.5=py_0
  - google-auth=1.11.2=py_0
  - google-auth-oauthlib=0.4.1=py_2
  - google-pasta=0.2.0=py_0
  - grpcio=1.27.2=py37hf8bcb03_0
  - gunicorn=20.0.4=py37_0
  - h5py=2.10.0=py37h7918eee_0
  - hdf5=1.10.4=hb1b8bf9_0
  - icu=58.2=he6710b0_3
  - idna=2.8=py37_0
  - intel-openmp=2020.0=166
  - ipykernel=5.1.4=py37h39e3cac_0
  - ipython=7.12.0=py37h5ca1d4c_0
  - ipython_genutils=0.2.0=py37_0
  - itsdangerous=1.1.0=py37_0
  - jedi=0.14.1=py37_0
  - jinja2=2.11.1=py_0
  - jmespath=0.9.4=py_0
  - joblib=0.14.1=py_0
  - jpeg=9b=h024ee3a_2
  - jupyter_client=5.3.4=py37_0
  - jupyter_core=4.6.1=py37_0
  - kiwisolver=1.1.0=py37he6710b0_0
  - krb5=1.16.4=h173b8e3_0
  - ld_impl_linux-64=2.33.1=h53a641e_7
  - libedit=3.1.20181209=hc058e9b_0
  - libffi=3.2.1=hd88cf55_4
  - libgcc-ng=9.1.0=hdf63c60_0
  - libgfortran-ng=7.3.0=hdf63c60_0
  - libpng=1.6.37=hbc83047_0
  - libpq=11.2=h20c2e04_0
  - libprotobuf=3.11.4=hd408876_0
  - libsodium=1.0.16=h1bed415_0
  - libstdcxx-ng=9.1.0=hdf63c60_0
  - libtiff=4.1.0=h2733197_0
  - lightgbm=2.3.0=py37he6710b0_0
  - lz4-c=1.8.1.2=h14c3975_0
  - mako=1.1.2=py_0
  - markdown=3.1.1=py37_0
  - markupsafe=1.1.1=py37h7b6447c_0
  - matplotlib-base=3.1.3=py37hef1b27d_0
  - mkl=2020.0=166
  - mkl-service=2.3.0=py37he904b0f_0
  - mkl_fft=1.0.15=py37ha843d7b_0
  - mkl_random=1.1.0=py37hd6b4f25_0
  - ncurses=6.2=he6710b0_1
  - networkx=2.4=py_0
  - ninja=1.9.0=py37hfd86e86_0
  - nltk=3.4.5=py37_0
  - numpy=1.18.1=py37h4f9e942_0
  - numpy-base=1.18.1=py37hde5b4d6_1
  - oauthlib=3.1.0=py_0
  - olefile=0.46=py37_0
  - openssl=1.1.1g=h7b6447c_0
  - packaging=20.1=py_0
  - pandas=1.0.1=py37h0573a6f_0
  - paramiko=2.7.1=py_0
  - parso=0.5.2=py_0
  - patsy=0.5.1=py37_0
  - pexpect=4.8.0=py37_0
  - pickleshare=0.7.5=py37_0
  - pillow=7.0.0=py37hb39fc2d_0
  - pip=20.0.2=py37_3
  - plotly=4.5.2=py_0
  - prompt_toolkit=3.0.3=py_0
  - protobuf=3.11.4=py37he6710b0_0
  - psutil=5.6.7=py37h7b6447c_0
  - psycopg2=2.8.4=py37h1ba5d50_0
  - ptyprocess=0.6.0=py37_0
  - pyasn1=0.4.8=py_0
  - pyasn1-modules=0.2.7=py_0
  - pycparser=2.19=py37_0
  - pygments=2.5.2=py_0
  - pyjwt=1.7.1=py37_0
  - pynacl=1.3.0=py37h7b6447c_0
  - pyodbc=4.0.30=py37he6710b0_0
  - pyopenssl=19.1.0=py37_0
  - pyparsing=2.4.6=py_0
  - pysocks=1.7.1=py37_0
  - python=3.7.6=h0371630_2
  - python-dateutil=2.8.1=py_0
  - python-editor=1.0.4=py_0
  - pytorch=1.5.0=py3.7_cuda10.1.243_cudnn7.6.3_0
  - pytz=2019.3=py_0
  - pyzmq=18.1.1=py37he6710b0_0
  - readline=7.0=h7b6447c_5
  - requests=2.22.0=py37_1
  - requests-oauthlib=1.3.0=py_0
  - retrying=1.3.3=py37_2
  - rsa=4.0=py_0
  - s3transfer=0.3.3=py37_0
  - scikit-learn=0.22.1=py37hd81dba3_0
  - scipy=1.4.1=py37h0b6359f_0
  - setuptools=45.2.0=py37_0
  - simplejson=3.17.0=py37h7b6447c_0
  - six=1.14.0=py37_0
  - smmap2=2.0.5=py37_0
  - sqlite=3.31.1=h62c20be_1
  - sqlparse=0.3.0=py_0
  - statsmodels=0.11.0=py37h7b6447c_0
  - tabulate=0.8.3=py37_0
  - tk=8.6.8=hbc83047_0
  - torchvision=0.6.0=py37_cu101
  - tornado=6.0.3=py37h7b6447c_3
  - tqdm=4.42.1=py_0
  - traitlets=4.3.3=py37_0
  - unixodbc=2.3.7=h14c3975_0
  - urllib3=1.25.8=py37_0
  - wcwidth=0.1.8=py_0
  - websocket-client=0.56.0=py37_0
  - werkzeug=1.0.0=py_0
  - wheel=0.34.2=py37_0
  - wrapt=1.11.2=py37h7b6447c_0
  - xz=5.2.4=h14c3975_4
  - zeromq=4.3.1=he6710b0_3
  - zlib=1.2.11=h7b6447c_3
  - zstd=1.3.7=h0b5b093_0
  - pip:
    - astunparse==1.6.3
    - databricks-cli==0.11.0
    - diskcache==4.1.0
    - docker==4.2.1
    - gorilla==0.3.0
    - horovod==0.19.1
    - hyperopt==0.2.4.db1
    - keras-preprocessing==1.1.2
    - mleap==0.16.0
    - mlflow==1.8.0
    - opt-einsum==3.2.1
    - petastorm==0.9.2
    - pyarrow==0.15.1
    - pyyaml==5.3.1
    - querystring-parser==1.2.4
    - seaborn==0.10.0
    - sparkdl==2.1.0-db1
    - tensorboard==2.2.2
    - tensorboard-plugin-wit==1.6.0.post3
    - tensorflow-estimator==2.2.0
    - tensorflow-gpu==2.2.0
    - termcolor==1.1.0
    - xgboost==1.1.1
prefix: /databricks/conda/envs/databricks-ml-gpu

Packages Spark contenant des modules Python

Package Spark Module Python Version
graphframes graphframes 0.8.0-db2-spark3.0

Bibliothèques R

Les bibliothèques R sont identiques aux bibliothèques R fournies dans Databricks Runtime 7.0 Bêta.

Bibliothèques Java et Scala (cluster Scala 2.12)

En plus des bibliothèques Java et Scala de Databricks Runtime 7.0, Databricks Runtime 7.0 ML contient les fichiers JAR suivants :

ID de groupe ID d’artefact Version
com.typesafe.akka akka-actor_2.12 2.5.23
ml.combust.mleap mleap-databricks-runtime_2.12 0.17.0-4882dc3
ml.dmlc xgboost4j-spark_2.12 1.0.0
ml.dmlc xgboost4j_2.12 1.0.0
org.mlflow mlflow-client 1.8.0
org.scala-lang.modules scala-java8-compat_2.12 0.8.0
org.tensorflow spark-tensorflow-connector_2.12 1.15.0