From de69b0d6d625615a58ac5a80d25f0cf81882b0f4 Mon Sep 17 00:00:00 2001 From: Kai Fricke Date: Sun, 15 May 2022 05:20:45 +0100 Subject: [PATCH] [train/release] Fix horovod user test master app config (#24734) --- .../horovod/app_config_master.yaml | 18 ++++++++++++++++++ release/release_tests.yaml | 2 +- 2 files changed, 19 insertions(+), 1 deletion(-) create mode 100644 release/ml_user_tests/horovod/app_config_master.yaml diff --git a/release/ml_user_tests/horovod/app_config_master.yaml b/release/ml_user_tests/horovod/app_config_master.yaml new file mode 100644 index 000000000..d518dd5a7 --- /dev/null +++ b/release/ml_user_tests/horovod/app_config_master.yaml @@ -0,0 +1,18 @@ +base_image: "anyscale/ray-ml:nightly-py37-gpu" +env_vars: {} +debian_packages: + - curl + +python: + pip_packages: + - pytest + - awscli + conda_packages: [] + +post_build_cmds: + - pip3 uninstall ray -y || true + - pip3 install -U {{ env["RAY_WHEELS"] | default("ray") }} + - pip3 install 'ray[tune]' + - pip3 install torch torchvision + - HOROVOD_WITH_GLOO=1 HOROVOD_WITHOUT_MPI=1 HOROVOD_WITHOUT_TENSORFLOW=1 HOROVOD_WITHOUT_MXNET=1 HOROVOD_WITH_PYTORCH=1 pip3 install -U git+https://github.com/horovod/horovod.git + - {{ env["RAY_WHEELS_SANITY_CHECK"] | default("echo No Ray wheels sanity check") }} diff --git a/release/release_tests.yaml b/release/release_tests.yaml index 4a3de9793..359b03245 100644 --- a/release/release_tests.yaml +++ b/release/release_tests.yaml @@ -593,7 +593,7 @@ team: ml cluster: - cluster_env: ../horovod_tests/app_config_master.yaml + cluster_env: horovod/app_config_master.yaml cluster_compute: horovod/compute_tpl.yaml driver_setup: horovod/driver_setup_master.sh