Creare un container Docker con la libreria SageMaker AI Distributed Data Parallelism
Per creare il tuo container Docker per l’addestramento e utilizzare la libreria SageMaker AI Distributed Data Parallelism, devi includere le dipendenze corrette e i file binari delle librerie SageMaker AI Distributed Data Parallelism nel tuo Dockerfile. Questa sezione fornisce istruzioni su come creare un Dockerfile completo con il set minimo di dipendenze per l’addestramento distribuito in SageMaker AI utilizzando la libreria SageMaker AI Distributed Data Parallelism.
Nota
Questa opzione del Docker personalizzata con la libreria SageMaker AI Distributed Data Parallelism come binario è disponibile solo per PyTorch.
Per creare un Dockerfile con il kit di strumenti di addestramento SageMaker e la libreria parallela dei dati
-
Inizia con un'immagine Docker di NVIDIA CUDA
. Usa le versioni per sviluppatori cuDNN che contengono runtime CUDA e strumenti di sviluppo (intestazioni e librerie) per creare dal codice sorgente PyTorch . FROM nvidia/cuda:11.3.1-cudnn8-devel-ubuntu20.04Suggerimento
Le immagini ufficiali del container AWS per il Deep Learning (DLC) sono create a partire dalle immagini di base di NVIDIA CUDA
. Se desideri utilizzare le immagini DLC predefinite come riferimenti mentre segui il resto delle istruzioni, consulta AWS Deep Learning Containers per Dockerfile PyTorch . -
Aggiungi i seguenti argomenti per specificare le versioni di PyTorch e altri pacchetti. Inoltre, indica i percorsi dei bucket Amazon S3 alla libreria SageMaker AI Distributed Data Parallelism e ad altri software per utilizzare le risorse AWS, come il plug-in Amazon S3.
Per utilizzare versioni delle librerie di terze parti diverse da quelle fornite nel seguente esempio di codice, ti consigliamo di esaminare i Dockerfile ufficiali di AWS Deep Learning Container per PyTorch
per trovare versioni testate, compatibili e adatte alla tua applicazione. Per trovare gli URL relativi all’argomento
SMDATAPARALLEL_BINARY, consulta le tabelle di ricerca in Framework supportati.ARG PYTORCH_VERSION=1.10.2ARG PYTHON_SHORT_VERSION=3.8ARG EFA_VERSION=1.14.1ARG SMDATAPARALLEL_BINARY=https://smdataparallel.s3.amazonaws.com/binary/pytorch/${PYTORCH_VERSION}/cu113/2022-02-18/smdistributed_dataparallel-1.4.0-cp38-cp38-linux_x86_64.whlARG PT_S3_WHL_GPU=https://aws-s3-plugin.s3.us-west-2.amazonaws.com/binaries/0.0.1/1c3e69e/awsio-0.0.1-cp38-cp38-manylinux1_x86_64.whl ARG CONDA_PREFIX="/opt/conda" ARG BRANCH_OFI=1.1.3-aws -
Imposta le seguenti variabili di ambiente per creare correttamente i componenti di addestramento SageMaker ed eseguire la libreria parallela dei dati. Queste variabili vengono utilizzate per i componenti nelle fasi successive.
# Set ENV variables required to build PyTorch ENV TORCH_CUDA_ARCH_LIST="7.0+PTX 8.0" ENV TORCH_NVCC_FLAGS="-Xfatbin -compress-all" ENV NCCL_VERSION=2.10.3 # Add OpenMPI to the path. ENV PATH /opt/amazon/openmpi/bin:$PATH # Add Conda to path ENV PATH $CONDA_PREFIX/bin:$PATH # Set this enviroment variable for SageMaker AI to launch SMDDP correctly. ENV SAGEMAKER_TRAINING_MODULE=sagemaker_pytorch_container.training:main # Add enviroment variable for processes to be able to call fork() ENV RDMAV_FORK_SAFE=1 # Indicate the container type ENV DLC_CONTAINER_TYPE=training # Add EFA and SMDDP to LD library path ENV LD_LIBRARY_PATH="/opt/conda/lib/python${PYTHON_SHORT_VERSION}/site-packages/smdistributed/dataparallel/lib:$LD_LIBRARY_PATH" ENV LD_LIBRARY_PATH=/opt/amazon/efa/lib/:$LD_LIBRARY_PATH -
Installa o aggiorna
curl,wgetegitper scaricare e creare pacchetti nelle fasi successive.RUN --mount=type=cache,id=apt-final,target=/var/cache/apt \ apt-get update && apt-get install -y --no-install-recommends \ curl \ wget \ git \ && rm -rf /var/lib/apt/lists/* -
Installa il software Elastic Fabric Adapter (EFA) per la comunicazione di rete Amazon EC2.
RUN DEBIAN_FRONTEND=noninteractive apt-get update RUN mkdir /tmp/efa \ && cd /tmp/efa \ && curl --silent -O https://efa-installer.amazonaws.com/aws-efa-installer-${EFA_VERSION}.tar.gz \ && tar -xf aws-efa-installer-${EFA_VERSION}.tar.gz \ && cd aws-efa-installer \ && ./efa_installer.sh -y --skip-kmod -g \ && rm -rf /tmp/efa -
Installa Conda
per eseguire la gestione dei pacchetti. RUN curl -fsSL -v -o ~/miniconda.sh -O https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh && \ chmod +x ~/miniconda.sh && \ ~/miniconda.sh -b -p $CONDA_PREFIX && \ rm ~/miniconda.sh && \ $CONDA_PREFIX/bin/conda install -y python=${PYTHON_SHORT_VERSION} conda-build pyyaml numpy ipython && \ $CONDA_PREFIX/bin/conda clean -ya -
Ottieni, crea e installa PyTorch e le sue dipendenze. Creiamo PyTorch dal codice sorgente
perché dobbiamo avere il controllo della versione NCCL per garantire la compatibilità con il plug-in AWS OFI NCCL . -
Seguendo le fasi descritte in dockerfile ufficiale di PyTorch
, installa dipendenze di compilazione e configura ccache per accelerare la ricompilazione. RUN DEBIAN_FRONTEND=noninteractive \ apt-get install -y --no-install-recommends \ build-essential \ ca-certificates \ ccache \ cmake \ git \ libjpeg-dev \ libpng-dev \ && rm -rf /var/lib/apt/lists/* # Setup ccache RUN /usr/sbin/update-ccache-symlinks RUN mkdir /opt/ccache && ccache --set-config=cache_dir=/opt/ccache -
Installa le dipendenze comuni e Linux di PyTorch
. # Common dependencies for PyTorch RUN conda install astunparse numpy ninja pyyaml mkl mkl-include setuptools cmake cffi typing_extensions future six requests dataclasses # Linux specific dependency for PyTorch RUN conda install -c pytorch magma-cuda113 -
Clona il repository PyTorch GitHub
. RUN --mount=type=cache,target=/opt/ccache \ cd / \ && git clone --recursive https://github.com/pytorch/pytorch -b v${PYTORCH_VERSION} -
Installa e crea una versione NCCL
specifica. Per fare ciò, sostituisci il contenuto nella cartella NCCL predefinita di PyTorch ( /pytorch/third_party/nccl) con la versione NCCL specifica del repository NVIDIA. La versione NCCL è stata impostata nella fase 3 di questa guida.RUN cd /pytorch/third_party/nccl \ && rm -rf nccl \ && git clone https://github.com/NVIDIA/nccl.git -b v${NCCL_VERSION}-1 \ && cd nccl \ && make -j64 src.build CUDA_HOME=/usr/local/cuda NVCC_GENCODE="-gencode=arch=compute_70,code=sm_70 -gencode=arch=compute_80,code=sm_80" \ && make pkg.txz.build \ && tar -xvf build/pkg/txz/nccl_*.txz -C $CONDA_PREFIX --strip-components=1 -
Compila e installa PyTorch. Questo processo dura in genere poco più di 1 ora. È costruito utilizzando la versione NCCL scaricata nella fase precedente.
RUN cd /pytorch \ && CMAKE_PREFIX_PATH="$(dirname $(which conda))/../" \ python setup.py install \ && rm -rf /pytorch
-
-
Crea e installa il plugin AWS OFI NCCL
. Questo abilita il supporto libfabric per la libreria SageMaker AI Distributed Data Parallelism. RUN DEBIAN_FRONTEND=noninteractive apt-get update \ && apt-get install -y --no-install-recommends \ autoconf \ automake \ libtool RUN mkdir /tmp/efa-ofi-nccl \ && cd /tmp/efa-ofi-nccl \ && git clone https://github.com/aws/aws-ofi-nccl.git -b v${BRANCH_OFI} \ && cd aws-ofi-nccl \ && ./autogen.sh \ && ./configure --with-libfabric=/opt/amazon/efa \ --with-mpi=/opt/amazon/openmpi \ --with-cuda=/usr/local/cuda \ --with-nccl=$CONDA_PREFIX \ && make \ && make install \ && rm -rf /tmp/efa-ofi-nccl -
Compila e installa TorchVision
. RUN pip install --no-cache-dir -U \ packaging \ mpi4py==3.0.3 RUN cd /tmp \ && git clone https://github.com/pytorch/vision.git -b v0.9.1 \ && cd vision \ && BUILD_VERSION="0.9.1+cu111" python setup.py install \ && cd /tmp \ && rm -rf vision -
Installa e configura OpenSSH. OpenSSH è necessario per consentire a MPI di comunicare tra container. Consenti a OpenSSH di comunicare con i container senza chiedere conferma.
RUN apt-get update \ && apt-get install -y --allow-downgrades --allow-change-held-packages --no-install-recommends \ && apt-get install -y --no-install-recommends openssh-client openssh-server \ && mkdir -p /var/run/sshd \ && cat /etc/ssh/ssh_config | grep -v StrictHostKeyChecking > /etc/ssh/ssh_config.new \ && echo " StrictHostKeyChecking no" >> /etc/ssh/ssh_config.new \ && mv /etc/ssh/ssh_config.new /etc/ssh/ssh_config \ && rm -rf /var/lib/apt/lists/* # Configure OpenSSH so that nodes can communicate with each other RUN mkdir -p /var/run/sshd && \ sed 's@session\s*required\s*pam_loginuid.so@session optional pam_loginuid.so@g' -i /etc/pam.d/sshd RUN rm -rf /root/.ssh/ && \ mkdir -p /root/.ssh/ && \ ssh-keygen -q -t rsa -N '' -f /root/.ssh/id_rsa && \ cp /root/.ssh/id_rsa.pub /root/.ssh/authorized_keys \ && printf "Host *\n StrictHostKeyChecking no\n" >> /root/.ssh/config -
Installa il plug-in PT S3 per accedere in modo efficiente ai set di dati in Amazon S3.
RUN pip install --no-cache-dir -U ${PT_S3_WHL_GPU} RUN mkdir -p /etc/pki/tls/certs && cp /etc/ssl/certs/ca-certificates.crt /etc/pki/tls/certs/ca-bundle.crt -
Installa la libreria libboost
. Questo pacchetto è necessario per collegare in rete la funzionalità IO asincrona della libreria SageMaker AI Distributed Data Parallelism. WORKDIR / RUN wget https://sourceforge.net/projects/boost/files/boost/1.73.0/boost_1_73_0.tar.gz/download -O boost_1_73_0.tar.gz \ && tar -xzf boost_1_73_0.tar.gz \ && cd boost_1_73_0 \ && ./bootstrap.sh \ && ./b2 threading=multi --prefix=${CONDA_PREFIX} -j 64 cxxflags=-fPIC cflags=-fPIC install || true \ && cd .. \ && rm -rf boost_1_73_0.tar.gz \ && rm -rf boost_1_73_0 \ && cd ${CONDA_PREFIX}/include/boost -
Installa i seguenti strumenti SageMaker AI per l’addestramento PyTorch.
WORKDIR /root RUN pip install --no-cache-dir -U \ smclarify \ "sagemaker>=2,<3" \ sagemaker-experiments==0.* \ sagemaker-pytorch-training -
Infine, installa il binario di parallelizzazione dei dati di SageMaker AI e le dipendenze rimanenti.
RUN --mount=type=cache,id=apt-final,target=/var/cache/apt \ apt-get update && apt-get install -y --no-install-recommends \ jq \ libhwloc-dev \ libnuma1 \ libnuma-dev \ libssl1.1 \ libtool \ hwloc \ && rm -rf /var/lib/apt/lists/* RUN SMDATAPARALLEL_PT=1 pip install --no-cache-dir ${SMDATAPARALLEL_BINARY} -
Dopo aver finito di creare il Dockerfile, consulta Adattamento del tuo container di addestramento per scoprire come creare il container Docker, ospitarlo in Amazon ECR ed eseguire un processo di addestramento utilizzando SageMaker Python SDK.
Il codice di esempio seguente mostra un Dockerfile completo dopo aver combinato tutti i blocchi di codice precedenti.
# This file creates a docker image with minimum dependencies to run SageMaker AI data parallel training FROM nvidia/cuda:11.3.1-cudnn8-devel-ubuntu20.04 # Set appropiate versions and location for components ARG PYTORCH_VERSION=1.10.2 ARG PYTHON_SHORT_VERSION=3.8 ARG EFA_VERSION=1.14.1 ARG SMDATAPARALLEL_BINARY=https://smdataparallel.s3.amazonaws.com/binary/pytorch/${PYTORCH_VERSION}/cu113/2022-02-18/smdistributed_dataparallel-1.4.0-cp38-cp38-linux_x86_64.whl ARG PT_S3_WHL_GPU=https://aws-s3-plugin.s3.us-west-2.amazonaws.com/binaries/0.0.1/1c3e69e/awsio-0.0.1-cp38-cp38-manylinux1_x86_64.whl ARG CONDA_PREFIX="/opt/conda" ARG BRANCH_OFI=1.1.3-aws # Set ENV variables required to build PyTorch ENV TORCH_CUDA_ARCH_LIST="3.7 5.0 7.0+PTX 8.0" ENV TORCH_NVCC_FLAGS="-Xfatbin -compress-all" ENV NCCL_VERSION=2.10.3 # Add OpenMPI to the path. ENV PATH /opt/amazon/openmpi/bin:$PATH # Add Conda to path ENV PATH $CONDA_PREFIX/bin:$PATH # Set this enviroment variable for SageMaker AI to launch SMDDP correctly. ENV SAGEMAKER_TRAINING_MODULE=sagemaker_pytorch_container.training:main # Add enviroment variable for processes to be able to call fork() ENV RDMAV_FORK_SAFE=1 # Indicate the container type ENV DLC_CONTAINER_TYPE=training # Add EFA and SMDDP to LD library path ENV LD_LIBRARY_PATH="/opt/conda/lib/python${PYTHON_SHORT_VERSION}/site-packages/smdistributed/dataparallel/lib:$LD_LIBRARY_PATH" ENV LD_LIBRARY_PATH=/opt/amazon/efa/lib/:$LD_LIBRARY_PATH # Install basic dependencies to download and build other dependencies RUN --mount=type=cache,id=apt-final,target=/var/cache/apt \ apt-get update && apt-get install -y --no-install-recommends \ curl \ wget \ git \ && rm -rf /var/lib/apt/lists/* # Install EFA. # This is required for SMDDP backend communication RUN DEBIAN_FRONTEND=noninteractive apt-get update RUN mkdir /tmp/efa \ && cd /tmp/efa \ && curl --silent -O https://efa-installer.amazonaws.com/aws-efa-installer-${EFA_VERSION}.tar.gz \ && tar -xf aws-efa-installer-${EFA_VERSION}.tar.gz \ && cd aws-efa-installer \ && ./efa_installer.sh -y --skip-kmod -g \ && rm -rf /tmp/efa # Install Conda RUN curl -fsSL -v -o ~/miniconda.sh -O https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh && \ chmod +x ~/miniconda.sh && \ ~/miniconda.sh -b -p $CONDA_PREFIX && \ rm ~/miniconda.sh && \ $CONDA_PREFIX/bin/conda install -y python=${PYTHON_SHORT_VERSION} conda-build pyyaml numpy ipython && \ $CONDA_PREFIX/bin/conda clean -ya # Install PyTorch. # Start with dependencies listed in official PyTorch dockerfile # https://github.com/pytorch/pytorch/blob/master/Dockerfile RUN DEBIAN_FRONTEND=noninteractive \ apt-get install -y --no-install-recommends \ build-essential \ ca-certificates \ ccache \ cmake \ git \ libjpeg-dev \ libpng-dev && \ rm -rf /var/lib/apt/lists/* # Setup ccache RUN /usr/sbin/update-ccache-symlinks RUN mkdir /opt/ccache && ccache --set-config=cache_dir=/opt/ccache # Common dependencies for PyTorch RUN conda install astunparse numpy ninja pyyaml mkl mkl-include setuptools cmake cffi typing_extensions future six requests dataclasses # Linux specific dependency for PyTorch RUN conda install -c pytorch magma-cuda113 # Clone PyTorch RUN --mount=type=cache,target=/opt/ccache \ cd / \ && git clone --recursive https://github.com/pytorch/pytorch -b v${PYTORCH_VERSION} # Note that we need to use the same NCCL version for PyTorch and OFI plugin. # To enforce that, install NCCL from source before building PT and OFI plugin. # Install NCCL. # Required for building OFI plugin (OFI requires NCCL's header files and library) RUN cd /pytorch/third_party/nccl \ && rm -rf nccl \ && git clone https://github.com/NVIDIA/nccl.git -b v${NCCL_VERSION}-1 \ && cd nccl \ && make -j64 src.build CUDA_HOME=/usr/local/cuda NVCC_GENCODE="-gencode=arch=compute_70,code=sm_70 -gencode=arch=compute_80,code=sm_80" \ && make pkg.txz.build \ && tar -xvf build/pkg/txz/nccl_*.txz -C $CONDA_PREFIX --strip-components=1 # Build and install PyTorch. RUN cd /pytorch \ && CMAKE_PREFIX_PATH="$(dirname $(which conda))/../" \ python setup.py install \ && rm -rf /pytorch RUN ccache -C # Build and install OFI plugin. \ # It is required to use libfabric. RUN DEBIAN_FRONTEND=noninteractive apt-get update \ && apt-get install -y --no-install-recommends \ autoconf \ automake \ libtool RUN mkdir /tmp/efa-ofi-nccl \ && cd /tmp/efa-ofi-nccl \ && git clone https://github.com/aws/aws-ofi-nccl.git -b v${BRANCH_OFI} \ && cd aws-ofi-nccl \ && ./autogen.sh \ && ./configure --with-libfabric=/opt/amazon/efa \ --with-mpi=/opt/amazon/openmpi \ --with-cuda=/usr/local/cuda \ --with-nccl=$CONDA_PREFIX \ && make \ && make install \ && rm -rf /tmp/efa-ofi-nccl # Build and install Torchvision RUN pip install --no-cache-dir -U \ packaging \ mpi4py==3.0.3 RUN cd /tmp \ && git clone https://github.com/pytorch/vision.git -b v0.9.1 \ && cd vision \ && BUILD_VERSION="0.9.1+cu111" python setup.py install \ && cd /tmp \ && rm -rf vision # Install OpenSSH. # Required for MPI to communicate between containers, allow OpenSSH to talk to containers without asking for confirmation RUN apt-get update \ && apt-get install -y --allow-downgrades --allow-change-held-packages --no-install-recommends \ && apt-get install -y --no-install-recommends openssh-client openssh-server \ && mkdir -p /var/run/sshd \ && cat /etc/ssh/ssh_config | grep -v StrictHostKeyChecking > /etc/ssh/ssh_config.new \ && echo " StrictHostKeyChecking no" >> /etc/ssh/ssh_config.new \ && mv /etc/ssh/ssh_config.new /etc/ssh/ssh_config \ && rm -rf /var/lib/apt/lists/* # Configure OpenSSH so that nodes can communicate with each other RUN mkdir -p /var/run/sshd && \ sed 's@session\s*required\s*pam_loginuid.so@session optional pam_loginuid.so@g' -i /etc/pam.d/sshd RUN rm -rf /root/.ssh/ && \ mkdir -p /root/.ssh/ && \ ssh-keygen -q -t rsa -N '' -f /root/.ssh/id_rsa && \ cp /root/.ssh/id_rsa.pub /root/.ssh/authorized_keys \ && printf "Host *\n StrictHostKeyChecking no\n" >> /root/.ssh/config # Install PT S3 plugin. # Required to efficiently access datasets in Amazon S3 RUN pip install --no-cache-dir -U ${PT_S3_WHL_GPU} RUN mkdir -p /etc/pki/tls/certs && cp /etc/ssl/certs/ca-certificates.crt /etc/pki/tls/certs/ca-bundle.crt # Install libboost from source. # This package is needed for smdataparallel functionality (for networking asynchronous IO). WORKDIR / RUN wget https://sourceforge.net/projects/boost/files/boost/1.73.0/boost_1_73_0.tar.gz/download -O boost_1_73_0.tar.gz \ && tar -xzf boost_1_73_0.tar.gz \ && cd boost_1_73_0 \ && ./bootstrap.sh \ && ./b2 threading=multi --prefix=${CONDA_PREFIX} -j 64 cxxflags=-fPIC cflags=-fPIC install || true \ && cd .. \ && rm -rf boost_1_73_0.tar.gz \ && rm -rf boost_1_73_0 \ && cd ${CONDA_PREFIX}/include/boost # Install SageMaker AI PyTorch training. WORKDIR /root RUN pip install --no-cache-dir -U \ smclarify \ "sagemaker>=2,<3" \ sagemaker-experiments==0.* \ sagemaker-pytorch-training # Install SageMaker AI data parallel binary (SMDDP) # Start with dependencies RUN --mount=type=cache,id=apt-final,target=/var/cache/apt \ apt-get update && apt-get install -y --no-install-recommends \ jq \ libhwloc-dev \ libnuma1 \ libnuma-dev \ libssl1.1 \ libtool \ hwloc \ && rm -rf /var/lib/apt/lists/* # Install SMDDP RUN SMDATAPARALLEL_PT=1 pip install --no-cache-dir ${SMDATAPARALLEL_BINARY}
Suggerimento
Per informazioni più generali sulla creazione di un Dockerfile personalizzato per l’addestramento in SageMaker AI, consulta Utilizzo degli algoritmi di addestramento.
Suggerimento
Se desideri estendere il Dockerfile personalizzato per incorporare la libreria SageMaker AI Model Parallelism, consulta Creazione del proprio container Docker con la libreria di parallelismo dei modelli distribuiti SageMaker.