AWS Deep Learning AMI GPU PyTorch 1.9 (Amazon Linux 2)

Release Date: June 24, 2021
Created On: June 23, 2021
Last Updated: June 27, 2022

For help getting started, please see the AWS Deep Learning AMI Developer Guide.

AMI Name format: Deep Learning AMI GPU PyTorch 1.9.${PATCH_VERSION} (Amazon Linux 2) ${YYYY-MM-DD}

The AMI includes the following:

  • Supported AWS Service: EC2
  • Operating System: Amazon Linux 2
  • Compute Architecture: x86
  • Python: /usr/bin/python3.7
  • NVIDIA Driver: 510.47.03
  • NVIDIA CUDA11 stack:
    • CUDA, NCCL and cuDDN installation path: /usr/local/cuda-11.1/
  • EFA Installer: 1.15.1
  • EBS volume type: gp3
  • Supported EC2 Instances: G3, P3, P3dn, P4d, G5, G4dn
  • Query AMI-ID with AWSCLI (example region is us-east-1):
    • aws ec2 describe-images --region us-east-1 --owners amazon --filters 'Name=name,Values=Deep Learning AMI GPU PyTorch 1.9.? (Amazon Linux 2) ????????' 'Name=state,Values=available' --query 'reverse(sort_by(Images, &CreationDate))[:1].ImageId' --output text

Version 1.9.1

Release Date: 2022-06-22

Changed

Version 1.9.1

Release Date: 2021-10-08

Changed

  • For every instance launch using DLAMI, tag "aws-dlami-autogenerated-tag-do-not-delete" will be added which will allow AWS to collect instance type, instance ID, DLAMI type, and OS information. No information on the commands used within the DLAMI is collected or retained. No other information about the DLAMI is collected or retained. To opt out of usage tracking for your DLAMI, add a tag to your Amazon EC2 instance during launch. The tag should use the key OPT_OUT_TRACKING with the associated value set to true. For more information, see Tag your Amazon EC2 resources.
  • Updated Nvidia driver version to 460.91.03

Security

  • Updated docker version to docker-20.10.7-3
  • Updated Pytorch to 1.9.1

Version 1.9.0

Release Date: 2021-08-24

Changed

  • Updated jupyterlab to version 3.1.7, pillow to 8.3.1, and gym to 0.19.0.

 

Version 1.9.0

Release Date: 2021-06-24

Added

  • Installed OS level utility packages:
    • "chrony"
    • "git"
    • "mlocate"
    • "dkms"
    • "bzip2"
    • "bzip2-devel"
    • "autoconf"
    • "libtool"
    • "snappy"
    • "snappy-devel"
    • "libjpeg-turbo-devel"
    • "graphviz"
    • "libcurl-devel"
    • "openssl-devel"
    • "tmux"
    • "cmake"
    • "protobuf-devel"
    • "libtiff"
    • "emacs"
    • "dnsmasq"
    • "jq"
    • "libffi-devel"
    • "libxslt-devel"
    • "awslogs"
    • "libudev-devel"
    • "Development Tools"
  • Installed system-wide python utility packages:
    • "pip",
    • "wheel",
    • "setuptools",
    • "twine",
    • "tqdm",
    • "cmake",
    • "pipenv",
    • "pip-tools",
    • "pipdeptree",
    • "safety",
    • "psutil",
    • "nvidia-ml-py",
    • "scrapy",
    • "flask",
    • "dask",
    • "psycopg2-binary",
    • "SQLAlchemy",
    • "celery",
    • "httpie",
    • "aiohttp",
    • "click",
    • "requests",
    • "fastapi",
    • "black",
    • "mypy",
    • "yapf",
    • "pylint",
    • "pytest",
    • "boto3",
    • "awscli",
    • "ipython",
    • "chalice",
    • "Jinja2",
    • "redis".
  • Installed:
    • "nvidia-driver=460.32.03",
    • "fabric-manager=460.32.03",
    • "cuda=11.1.1",
    • "cudnn=8.0.5",
    • "nccl=2.7.8",
    • "aws-ofi-nccl",
    • "efa=1.12.1",
    • "docker",
    • "nvidia-docker",
    • "nvidia-persistenced".
  • Installed:
    • "java-11-amazon-corretto-headless"
  • created "pytorch" conda environment with below package plan:
    • Mambaforge-4.10.1-4-Linux-x86_64
    • conda packages:
      • python=3.7.10
      • pytorch=1.9.0=py3.7_cuda11.1_cudnn8.0.5_0
      • torchserve=0.4.*
      • torch-model-archiver=0.4.*
      • torch-workflow-archiver=0.1.*
      • captum=0.3.*
      • magma-cuda111=2.5.*
      • jupyterlab=3.0.*
      • imageio=2.9.*
      • pillow=8.2.*
      • matplotlib=3.4.*
      • plotly=4.14.*
      • pandas=1.2.*
      • bokeh=2.3.*
      • seaborn=0.11.*
      • psutil=5.8.*
      • cmake=3.20.*
      • pybind11=2.6.*
      • scipy=1.6.*
      • boto3=1.17.*
      • awscli=1.19.*
      • pip=21.1.*
    • pip packages:
      • blis==0.7.4
      • catalogue==2.0.4
      • click==7.1.2
      • cloudpickle==1.6.0
      • cymem==2.0.5
      • dill==0.3.4
      • dparse==0.5.1
      • fastai==2.1.10
      • fastcore==1.3.20
      • fastprogress==1.0.0
      • filelock==3.0.12
      • fsspec==2021.6.0
      • google-pasta==0.2.0
      • gym==0.18.3
      • joblib==1.0.1
      • jsonpatch==1.32
      • jsonpointer==2.1
      • llvmlite==0.36.0
      • lmdb==1.2.1
      • multiprocess==0.70.12.2
      • murmurhash==1.0.5
      • numba==0.53.1
      • nvidia-ml-py==11.460.79
      • opencv-python==4.5.2.54
      • pathos==0.2.8
      • pathy==0.5.2
      • pox==0.3.0
      • ppft==1.6.6.4
      • preshed==3.0.5
      • protobuf==3.17.3
      • protobuf3-to-dict==0.1.5
      • pyarrow==4.0.1
      • pydantic>=1.7.4
      • pyfunctional==1.4.3
      • pyglet==1.5.15
      • s3fs==0.4.2
      • safety==1.10.3
      • sagemaker>=2,<3
      • scikit-learn==0.24.2
      • shap==0.39.0
      • sklearn==0.0
      • slicer==0.0.7
      • smart-open==3.0.0
      • smclarify==0.2
      • smdebug-rulesconfig==1.0.1
      • spacy==3.0.6
      • spacy-legacy==3.0.6
      • srsly==2.4.1
      • tabulate==0.8.9
      • thinc>=8.0.4
      • threadpoolctl==2.1.0
      • toml==0.10.2
      • torchaudio==0.9.0
      • torchfile==0.1.0
      • torchnet==0.0.4
      • torchtext==0.10.0
      • torchvision=0.10.0-cu111
      • tqdm==4.61.1
      • typer==0.3.2
      • visdom==0.1.8.9
      • wasabi==0.8.2
      • horovod==0.22.1