From 5d28e925a85e79cb63db40d98b99fe15f1bc364c Mon Sep 17 00:00:00 2001 From: mhuguesaws <71357145+mhuguesaws@users.noreply.github.com> Date: Thu, 17 Oct 2024 08:42:54 -0500 Subject: [PATCH] Change EFA Installer and AWS OFI nccl plugin versions. (#454) --- micro-benchmarks/nccl-tests/README.md | 12 ++++++------ micro-benchmarks/nccl-tests/buildspec.yaml | 4 ++-- micro-benchmarks/nccl-tests/nccl-tests.Dockerfile | 4 ++-- 3 files changed, 10 insertions(+), 10 deletions(-) diff --git a/micro-benchmarks/nccl-tests/README.md b/micro-benchmarks/nccl-tests/README.md index d1aab3ad..bbd5ee9e 100644 --- a/micro-benchmarks/nccl-tests/README.md +++ b/micro-benchmarks/nccl-tests/README.md @@ -36,16 +36,16 @@ The NCCL tests are packaged in a container. > | Variable | Default | Repository | > |-----------------------|-------------|---------------------------------------------------------------------------------------------| > |`GDRCOPY_VERSION` | `v2.4.1` | [link](https://github.com/NVIDIA/gdrcopy) | -> |`EFA_INSTALLER_VERSION`| `1.34.0` | [link](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa-start.html#efa-start-enable) | -> |`AWS_OFI_NCCL_VERSION` | `v1.11.0-aws`| [link](https://github.com/aws/aws-ofi-nccl) | +> |`EFA_INSTALLER_VERSION`| `1.35.0` | [link](https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/efa-start.html#efa-start-enable) | +> |`AWS_OFI_NCCL_VERSION` | `v1.12.0-aws`| [link](https://github.com/aws/aws-ofi-nccl) | > |`NCCL_VERSION` | `v2.23.4-1` | [link](https://github.com/NVIDIA/nccl) | > |`NCCL_TESTS_VERSION` | `v2.13.10` | [link](https://github.com/NVIDIA/nccl-tests) | ### Build the container 1. Build the container image with the command below: ```bash - EFA_INSTALLER_VERSION=1.34.0 - AWS_OFI_NCCL_VERSION=v1.11.0-aws + EFA_INSTALLER_VERSION=1.35.0 + AWS_OFI_NCCL_VERSION=v1.12.0-aws NCCL_VERSION=v2.23.4-1 NCCL_TESTS_VERSION=v2.13.10 docker build -f nccl-tests.Dockerfile \ @@ -81,8 +81,8 @@ To run the NCCL tests on EKS, you will need to build the container image, then p 1. Create the ECR repository if it does not exist ```bash - EFA_INSTALLER_VERSION=1.34.0 - AWS_OFI_NCCL_VERSION=v1.11.0-aws + EFA_INSTALLER_VERSION=1.35.0 + AWS_OFI_NCCL_VERSION=v1.12.0-aws NCCL_VERSION=v2.23.4-1 NCCL_TESTS_VERSION=v2.13.10 ECR_REPOSITORY_NAME="nccl-tests" diff --git a/micro-benchmarks/nccl-tests/buildspec.yaml b/micro-benchmarks/nccl-tests/buildspec.yaml index 85082d22..8a064f1b 100644 --- a/micro-benchmarks/nccl-tests/buildspec.yaml +++ b/micro-benchmarks/nccl-tests/buildspec.yaml @@ -3,8 +3,8 @@ version: 0.2 env: variables: GDRCOPY_VERSION: "v2.4.1" - EFA_INSTALLER_VERSION: "1.34.0" - AWS_OFI_NCCL_VERSION: "v1.11.0-aws" + EFA_INSTALLER_VERSION: "1.35.0" + AWS_OFI_NCCL_VERSION: "v1.12.0-aws" NCCL_VERSION: "v2.23.4-1" NCCL_TESTS_VERSION: "v2.13.10" exported-variables: diff --git a/micro-benchmarks/nccl-tests/nccl-tests.Dockerfile b/micro-benchmarks/nccl-tests/nccl-tests.Dockerfile index 3a531bfc..4aa71c03 100644 --- a/micro-benchmarks/nccl-tests/nccl-tests.Dockerfile +++ b/micro-benchmarks/nccl-tests/nccl-tests.Dockerfile @@ -3,8 +3,8 @@ FROM nvidia/cuda:12.2.2-devel-ubuntu22.04 ARG GDRCOPY_VERSION=v2.4.1 -ARG EFA_INSTALLER_VERSION=1.34.0 -ARG AWS_OFI_NCCL_VERSION=v1.11.0-aws +ARG EFA_INSTALLER_VERSION=1.35.0 +ARG AWS_OFI_NCCL_VERSION=v1.12.0-aws ARG NCCL_VERSION=v2.23.4-1 ARG NCCL_TESTS_VERSION=v2.13.10