From b11805e6149b5c1b847ff5b657bb706fe7be71b4 Mon Sep 17 00:00:00 2001 From: Kai Fricke Date: Mon, 21 Jun 2021 15:19:00 +0100 Subject: [PATCH 1/3] Install latest releases, bump xgboost version --- python/requirements/tune/requirements_upstream.txt | 2 +- release/tune_tests/scalability_tests/app_config_data.yaml | 1 + release/xgboost_tests/app_config.yaml | 1 + 3 files changed, 3 insertions(+), 1 deletion(-) diff --git a/python/requirements/tune/requirements_upstream.txt b/python/requirements/tune/requirements_upstream.txt index bf2e0e017fa7e..b3d5ce0ef7406 100644 --- a/python/requirements/tune/requirements_upstream.txt +++ b/python/requirements/tune/requirements_upstream.txt @@ -3,5 +3,5 @@ # So we separate its own requirements file. tune-sklearn==0.3.0 -xgboost_ray==0.0.5 +xgboost_ray==0.1.1 modin>=0.10.0; python_version >= '3.7' \ No newline at end of file diff --git a/release/tune_tests/scalability_tests/app_config_data.yaml b/release/tune_tests/scalability_tests/app_config_data.yaml index 79d9bc31b851c..f0f2c6b983aac 100755 --- a/release/tune_tests/scalability_tests/app_config_data.yaml +++ b/release/tune_tests/scalability_tests/app_config_data.yaml @@ -15,6 +15,7 @@ post_build_cmds: - sudo rm -rf /home/ray/anaconda3/lib/python3.7/site-packages/numpy - pip3 install numpy || true - pip3 install -U {{ env["RAY_WHEELS"] | default("ray") }} + - pip3 install -U xgboost_ray awscli # Install latest releases - sudo mkdir -p /data || true - sudo chown ray:1000 /data || true - rm -rf /data/train.parquet || true diff --git a/release/xgboost_tests/app_config.yaml b/release/xgboost_tests/app_config.yaml index 2bd1b5c6fed88..ecd758735355b 100755 --- a/release/xgboost_tests/app_config.yaml +++ b/release/xgboost_tests/app_config.yaml @@ -15,6 +15,7 @@ post_build_cmds: - sudo rm -rf /home/ray/anaconda3/lib/python3.7/site-packages/numpy - pip3 install numpy || true - pip3 install -U {{ env["RAY_WHEELS"] | default("ray") }} + - pip3 install -U xgboost_ray petastorm # Install latest releases - sudo mkdir -p /data || true - sudo chown ray:1000 /data || true - rm -rf /data/classification.parquet || true From 345057f14f8cb4d982755340962498b5af6ad80f Mon Sep 17 00:00:00 2001 From: Kai Fricke Date: Tue, 22 Jun 2021 15:43:00 +0100 Subject: [PATCH 2/3] Horovod smoke test --- release/horovod_tests/horovod_tests.yaml | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/release/horovod_tests/horovod_tests.yaml b/release/horovod_tests/horovod_tests.yaml index 875bf898e0b16..b4b7fd4b72844 100644 --- a/release/horovod_tests/horovod_tests.yaml +++ b/release/horovod_tests/horovod_tests.yaml @@ -8,3 +8,7 @@ script: python workloads/horovod_test.py time_key: last_update max_update_delay: 120 + + smoke_test: + timeout: 3600 + long_running: True From c5c50fe44be56281b77eeb06b0bfdc52183cc2c3 Mon Sep 17 00:00:00 2001 From: Kai Fricke Date: Tue, 22 Jun 2021 15:47:21 +0100 Subject: [PATCH 3/3] Long running distributed --- release/horovod_tests/horovod_tests.yaml | 6 ++---- .../long_running_distributed.yaml | 8 +++++--- 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/release/horovod_tests/horovod_tests.yaml b/release/horovod_tests/horovod_tests.yaml index b4b7fd4b72844..c4be80090c03c 100644 --- a/release/horovod_tests/horovod_tests.yaml +++ b/release/horovod_tests/horovod_tests.yaml @@ -4,11 +4,9 @@ compute_template: compute_tpl.yaml run: - timeout: 90000 + timeout: 86400 script: python workloads/horovod_test.py - time_key: last_update - max_update_delay: 120 + long_running: True smoke_test: timeout: 3600 - long_running: True diff --git a/release/long_running_distributed_tests/long_running_distributed.yaml b/release/long_running_distributed_tests/long_running_distributed.yaml index 3e2ef67c9ba3e..132cb9ea3c72e 100644 --- a/release/long_running_distributed_tests/long_running_distributed.yaml +++ b/release/long_running_distributed_tests/long_running_distributed.yaml @@ -4,7 +4,9 @@ compute_template: compute_tpl.yaml run: - timeout: 90000 + timeout: 86400 script: python workloads/pytorch_pbt_failure.py - time_key: last_update - max_update_delay: 120 + long_running: True + + smoke_test: + timeout: 3600