# Global release test configuration file. # All your release test configuration should go here. Adding release tests here # will automatically enable them in the Buildkite release testing schedules # (except they have frequency: manual). # Here is an example configuration for reference: #- name: example_test # # Tests with the same group will be grouped in the Buildkite UI # group: Example group # # Provide the working directory which will be uploaded to the cluster # working_dir: example_dir # # # How often to run the tests. # # One of [manual, any, multi, nightly, nightly-3x, weekly]. # # Descriptions of each frequency (that's not immediately obvious): # # - manual: Not run on a schedule, but can be manually run through the buildkite UI. # # - nightly-3x: Run 3 times a week (Monday, Wednesday, Friday). # frequency: weekly # # Owning team. This field will be persisted to the database # team: ml # # # Python version. This optional field determines which Python version to run tests # # on. This must be a string! # python: "3.7" # # # Cluster information # cluster: # # Location of cluster env, relative to working_dir # cluster_env: cluster_env.yaml # # Location of cluster compute, relative to working_dir # cluster_compute: cluster_compute.yaml # # Autosuspend parameter passed to the cluster. # # The cluster will automatically terminate if inactive for this # # many minutes. Defaults to 10 if not set. # autosuspend_mins: 10 # # Optional cloud_id to use instead of the default cloud # cloud_id: cld_12345678 # # Alternatively, you can specify a cloud name # cloud_name: anyscale_default_cloud # # # Run configuration for the test # run: # # If you want to wait for nodes to be ready, you can specify this here: # wait_for_nodes: # # Number of nodes # num_nodes: 16 # # Timeout for waiting for nodes. If nodes are not up by then, the # # test will fail. # timeout: 600 # # # Optional prepare script to be run on the cluster before the test script # prepare: python prepare.py # # The prepare command can have a separate timeout # prepare_timeout: 300 # # # Main script to run as the test script # script: python workloads/train_small.py # # Timeout in seconds. After this time the test is considered as failed. # timeout: 600 # # # You can specify smoke test definitions here. If a smoke test is triggered, # # it will deep update the main test configuration with the values provided # # here. Smoke tests will automatically run with IS_SMOKE_TEST=1 as en # # environment variable and receive the --smoke-test flag as a parameter in the # # run script. # smoke_test: # # Smoke tests can have different frequencies. A smoke test is only triggered # # when the regular test is not matched. # frequency: nightly # # Here we adjust the run timeout down and run on less nodes. The test script # # remains the same. # run: # timeout: 300 # wait_for_nodes: # num_nodes: 4 # timeout: 600 # # # After the test finished, this handler (in alerts/) will process the results. # # It can then let the test fail, e.g. if a metric regression is observed. # alert: default ####################### # Cluster scaling tests ####################### - name: cluster_tune_scale_up_down group: Cluster tests working_dir: cluster_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: cpt_autoscaling_1-3_aws.yaml run: timeout: 3600 script: python workloads/tune_scale_up_down.py wait_for_nodes: num_nodes: 0 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: cpt_autoscaling_1-3_gce.yaml alert: default ######################### # AIR release tests ######################### - name: tune_with_frequent_pausing group: AIR tests working_dir: air_tests frequency: nightly-3x team: ml cluster: cluster_env: frequent_pausing/app_config.yaml cluster_compute: frequent_pausing/compute_config_aws.yaml run: timeout: 600 # 10min long_running: true script: python frequent_pausing/script.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: frequent_pausing/app_config.yaml cluster_compute: frequent_pausing/compute_config_gce.yaml alert: default - name: long_running_horovod_tune_test group: AIR tests working_dir: air_tests frequency: weekly team: ml cluster: cluster_env: horovod/app_config_master.yaml cluster_compute: horovod/compute_tpl_aws.yaml variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: horovod/app_config_master.yaml cluster_compute: horovod/compute_tpl_gce.yaml run: timeout: 36000 script: python horovod/workloads/horovod_tune_test.py long_running: true wait_for_nodes: num_nodes: 2 smoke_test: frequency: manual run: timeout: 3600 alert: default - name: air_benchmark_data_bulk_ingest group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_data_20_nodes_aws.yaml run: timeout: 3600 script: python workloads/data_benchmark.py --dataset-size-gb=200 --num-workers=20 wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_data_20_nodes_gce.yaml alert: default # AIR benchmarks for XGBoost CUJ - name: air_benchmark_xgboost_cpu_10 group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: xgboost_app_config.yaml cluster_compute: compute_xgboost_aws.yaml run: timeout: 36000 script: python workloads/xgboost_benchmark.py wait_for_nodes: num_nodes: 11 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: xgboost_app_config.yaml cluster_compute: compute_xgboost_gce.yaml smoke_test: frequency: manual run: timeout: 1800 alert: default # Ray AIR distributed Torch benchmarks - name: air_benchmark_torch_mnist_cpu_4x1 group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_4_aws.yaml run: timeout: 3600 script: python workloads/torch_benchmark.py run --num-runs 3 --num-epochs 20 --num-workers 4 --cpus-per-worker 8 wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_4_gce.yaml alert: default - name: air_benchmark_torch_mnist_gpu_4x4 group: AIR tests working_dir: air_tests/air_benchmarks frequency: weekly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_aws.yaml run: timeout: 4800 script: python workloads/torch_benchmark.py run --num-runs 3 --num-epochs 120 --num-workers 16 --cpus-per-worker 4 --batch-size 1024 --use-gpu wait_for_nodes: num_nodes: 4 smoke_test: frequency: nightly cluster: cluster_compute: compute_gpu_2x2_aws.yaml run: timeout: 3600 script: python workloads/torch_benchmark.py run --num-runs 3 --num-epochs 60 --num-workers 4 --cpus-per-worker 4 --batch-size 512 --use-gpu wait_for_nodes: num_nodes: 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_gce.yaml smoke_test: frequency: manual alert: default - name: air_benchmark_torch_mnist_cpu_1x4 group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_1_aws.yaml run: timeout: 3600 script: python workloads/torch_benchmark.py run --num-runs 3 --num-epochs 20 --num-workers 4 --cpus-per-worker 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_1_gce.yaml alert: default - name: air_benchmark_torch_batch_prediction_gpu_1x1_20gb group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_1_cpu_16_aws.yaml run: timeout: 3600 script: python workloads/gpu_batch_prediction.py --data-size-gb 20 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_1_cpu_16_gce.yaml - name: air_benchmark_torch_batch_prediction_gpu_4x4_100gb group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml stable: false cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_aws.yaml run: timeout: 10800 script: python workloads/gpu_batch_prediction.py --data-size-gb 100 wait_for_nodes: num_nodes: 4 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_gce.yaml - name: air_benchmark_torch_mnist_cpu_4x4 group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_4_aws.yaml run: timeout: 5400 script: python workloads/torch_benchmark.py run --num-runs 3 --num-epochs 20 --num-workers 16 --cpus-per-worker 2 wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_4_gce.yaml alert: default - name: air_benchmark_tune_torch_mnist group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_8_aws.yaml run: timeout: 3600 script: python workloads/tune_torch_benchmark.py --num-runs 3 --num-trials 8 --num-workers 4 wait_for_nodes: num_nodes: 8 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_8_gce.yaml alert: default - name: air_benchmark_tune_torch_mnist_gpu group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_aws.yaml run: timeout: 3600 script: python workloads/tune_torch_benchmark.py --num-runs 2 --num-trials 4 --num-workers 4 --use-gpu wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_gce.yaml alert: default # Ray AIR distributed Tensorflow benchmarks - name: air_benchmark_tensorflow_mnist_cpu_4x1 group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_4_aws.yaml run: timeout: 5400 script: python workloads/tensorflow_benchmark.py run --num-runs 3 --num-epochs 20 --num-workers 4 --cpus-per-worker 8 wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_4_gce.yaml alert: default - name: air_benchmark_tensorflow_mnist_cpu_1x4 group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_1_aws.yaml run: timeout: 5400 script: python workloads/tensorflow_benchmark.py run --num-runs 3 --num-epochs 20 --num-workers 4 --cpus-per-worker 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_1_gce.yaml alert: default - name: air_benchmark_tensorflow_mnist_cpu_4x4 group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml stable: false cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_4_aws.yaml run: timeout: 5400 script: python workloads/tensorflow_benchmark.py run --num-runs 3 --num-epochs 20 --num-workers 16 --cpus-per-worker 2 wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_cpu_4_gce.yaml alert: default - name: air_benchmark_tensorflow_mnist_gpu_4x4 group: AIR tests working_dir: air_tests/air_benchmarks frequency: weekly team: ml stable: false cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_aws.yaml run: timeout: 5400 script: python workloads/tensorflow_benchmark.py run --num-runs 3 --num-epochs 200 --num-workers 16 --cpus-per-worker 4 --batch-size 1024 --use-gpu wait_for_nodes: num_nodes: 4 smoke_test: frequency: nightly cluster: cluster_compute: compute_gpu_2x2_aws.yaml run: script: python workloads/tensorflow_benchmark.py run --num-runs 3 --num-epochs 60 --num-workers 4 --cpus-per-worker 4 --batch-size 512 --use-gpu wait_for_nodes: num_nodes: 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_gce.yaml smoke_test: frequency: manual alert: default - name: air_benchmark_pytorch_training_e2e_gpu_1x1_20gb group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_1_aws.yaml run: timeout: 3600 script: python workloads/pytorch_training_e2e.py --data-size-gb 20 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_1_gce.yaml - name: air_benchmark_pytorch_training_e2e_gpu_4x4_100gb group: AIR tests working_dir: air_tests/air_benchmarks frequency: nightly team: ml stable: false cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_aws.yaml run: timeout: 10800 script: python workloads/pytorch_training_e2e.py --data-size-gb=100 --num-workers=16 wait_for_nodes: num_nodes: 4 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_gpu_4x4_gce.yaml # Test tiny, medium, and huge input files. - name: ray-data-bulk-ingest-file-size-benchmark group: AIR tests working_dir: air_tests/air_benchmarks/mlperf-train stable: false jailed: true frequency: nightly team: data cluster: cluster_env: app_config_oom.yaml cluster_compute: compute_cpu_16.yaml run: timeout: 3600 script: bash file_size_benchmark.sh variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config_oom.yaml cluster_compute: compute_gce_cpu_16.yaml # Test dataset larger than object store memory. - name: ray-data-bulk-ingest-out-of-core-benchmark group: AIR tests working_dir: air_tests/air_benchmarks/mlperf-train stable: false jailed: true frequency: nightly team: data cluster: cluster_env: app_config_oom.yaml cluster_compute: compute_cpu_16.yaml run: timeout: 3600 script: bash out_of_core_benchmark.sh variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config_oom.yaml cluster_compute: compute_gce_cpu_16.yaml # Test additional CPU nodes for preprocessing. - name: ray-data-bulk-ingest-heterogeneity-benchmark group: AIR tests working_dir: air_tests/air_benchmarks/mlperf-train stable: false jailed: true frequency: nightly team: data cluster: cluster_env: app_config_oom.yaml cluster_compute: compute_cpu_16_worker_nodes_2.yaml run: wait_for_nodes: num_nodes: 3 timeout: 1800 script: bash heterogeneity_benchmark.sh 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config_oom.yaml cluster_compute: compute_gce_cpu_16_worker_nodes_2.yaml ####################### # AIR examples ####################### # Test additional CPU nodes for preprocessing. - name: air_example_dreambooth_finetuning group: AIR examples working_dir: air_examples/dreambooth stable: false frequency: weekly team: ml cluster: cluster_env: dreambooth_env.yaml cluster_compute: dreambooth_compute_aws.yaml run: timeout: 1800 script: bash dreambooth_run.sh artifact_path: /tmp/artifacts/example_out.jpg # variations: A10G not available on GCE, yet. - name: air_example_gptj_deepspeed_fine_tuning group: AIR examples working_dir: air_examples/gptj_deepspeed_finetuning python: "3.9" frequency: weekly team: ml cluster: cluster_env: gptj_deepspeed_env.yaml cluster_compute: gptj_deepspeed_compute_aws.yaml run: timeout: 3600 script: python test_myst_doc.py --path gptj_deepspeed_fine_tuning.ipynb variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: gptj_deepspeed_env.yaml cluster_compute: gptj_deepspeed_compute_gce.yaml - name: air_example_dolly_v2_lightning_fsdp_finetuning group: AIR examples working_dir: air_examples/dolly_v2_lightning_fsdp_finetuning python: "3.8" frequency: weekly team: ml cluster: cluster_env: dolly_v2_fsdp_env.yaml cluster_compute: dolly_v2_fsdp_compute_aws.yaml run: timeout: 4700 script: python test_myst_doc.py --path lightning-llm-finetuning-7b.ipynb - name: air_example_opt_deepspeed_batch_inference group: AIR examples working_dir: air_examples/opt_deepspeed_batch_inference python: "3.9" frequency: weekly team: ml cluster: cluster_env: 30b_deepspeed_env.yaml cluster_compute: 30b_deepspeed_compute.yaml run: timeout: 3600 script: python test_myst_doc.py --path opt_deepspeed_batch_inference.ipynb # variations: TODO(jungong): add GCP variation. ##################################### # Workspace templates release tests # ##################################### - name: workspace_template_batch_inference group: Workspace templates working_dir: workspace_templates/01_batch_inference python: "3.9" frequency: nightly-3x team: ml cluster: cluster_env: ../testing/cluster_envs/default_cluster_env_latest_ml_py39.yaml cluster_compute: ../testing/compute_configs/gpu/aws.yaml run: timeout: 600 script: jupyter nbconvert --to script --output _test start.ipynb && ipython _test.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: ../testing/cluster_envs/default_cluster_env_latest_ml_py39.yaml cluster_compute: ../testing/compute_configs/gpu/gce.yaml - name: workspace_template_many_model_training group: Workspace templates working_dir: workspace_templates/02_many_model_training python: "3.9" frequency: nightly-3x team: ml cluster: cluster_env: ../testing/cluster_envs/default_cluster_env_latest_ml_py39.yaml cluster_compute: ../testing/compute_configs/cpu/aws.yaml run: timeout: 600 script: pip install --user -r requirements.txt && jupyter nbconvert --to script --output _test start.ipynb && ipython _test.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: ../testing/cluster_envs/default_cluster_env_latest_ml_py39.yaml cluster_compute: ../testing/compute_configs/cpu/gce.yaml - name: workspace_template_serving_stable_diffusion group: Workspace templates working_dir: workspace_templates/03_serving_stable_diffusion python: "3.9" frequency: nightly-3x team: ml cluster: cluster_env: ../testing/cluster_envs/03_serving_stable_diffusion.yaml cluster_compute: ../testing/compute_configs/gpu/aws.yaml run: timeout: 600 script: jupyter nbconvert --to script --output _test start.ipynb && ipython _test.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: ../testing/cluster_envs/03_serving_stable_diffusion.yaml cluster_compute: ../testing/compute_configs/gpu/gce.yaml ####################### # XGBoost release tests ####################### # It seems like the consensus is that we can deprecate this test. # - name: xgboost_train_small # group: XGBoost # working_dir: xgboost_tests # frequency: nightly # team: ml # env: staging_v2 # cluster: # cluster_env: app_config.yaml # cluster_compute: tpl_cpu_small.yaml # run: # timeout: 600 # script: python workloads/train_small.py # wait_for_nodes: # num_nodes: 4 # alert: xgboost_tests - name: xgboost_train_moderate group: XGBoost working_dir: xgboost_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_aws.yaml run: timeout: 600 script: python workloads/train_moderate.py wait_for_nodes: num_nodes: 32 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_gce.yaml alert: xgboost_tests - name: xgboost_train_gpu group: XGBoost working_dir: xgboost_tests frequency: nightly team: ml cluster: cluster_env: app_config_gpu.yaml cluster_compute: tpl_gpu_small_aws.yaml run: timeout: 600 script: python workloads/train_gpu.py wait_for_nodes: num_nodes: 5 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config_gpu.yaml cluster_compute: tpl_gpu_small_gce.yaml alert: xgboost_tests - name: xgboost_distributed_api_test group: XGBoost working_dir: xgboost_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_aws.yaml run: timeout: 600 script: python workloads/distributed_api_test.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_gce.yaml alert: default - name: xgboost_ft_small_elastic group: XGBoost working_dir: xgboost_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_aws.yaml run: timeout: 900 script: python workloads/ft_small_elastic.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_gce.yaml alert: default - name: xgboost_ft_small_non_elastic group: XGBoost working_dir: xgboost_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_aws.yaml run: timeout: 900 script: python workloads/ft_small_non_elastic.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_gce.yaml alert: default - name: xgboost_tune_small group: XGBoost working_dir: xgboost_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_aws.yaml run: timeout: 600 script: python workloads/tune_small.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_gce.yaml alert: xgboost_tests - name: xgboost_tune_32x4 group: XGBoost working_dir: xgboost_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_aws.yaml run: timeout: 900 script: python workloads/tune_32x4.py wait_for_nodes: num_nodes: 32 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_gce.yaml alert: xgboost_tests - name: xgboost_tune_4x32 group: XGBoost working_dir: xgboost_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_aws.yaml run: timeout: 900 script: python workloads/tune_4x32.py wait_for_nodes: num_nodes: 32 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_gce.yaml alert: xgboost_tests ####################### # LightGBM tests ####################### # It seems like the consensus is that we can deprecate this test. # - name: lightgbm_train_small # group: LightGBM tests # working_dir: lightgbm_tests # frequency: nightly # team: ml # env: staging_v2 # cluster: # cluster_env: app_config.yaml # cluster_compute: tpl_cpu_small.yaml # run: # timeout: 600 # script: python workloads/train_small.py # wait_for_nodes: # num_nodes: 4 # alert: default - name: lightgbm_train_moderate group: LightGBM tests working_dir: lightgbm_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_aws.yaml run: timeout: 600 script: python workloads/train_moderate.py wait_for_nodes: num_nodes: 32 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_gce.yaml alert: default - name: lightgbm_distributed_api_test group: LightGBM tests working_dir: lightgbm_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_aws.yaml run: timeout: 600 script: python workloads/distributed_api_test.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_gce.yaml alert: default - name: lightgbm_ft_small_non_elastic group: LightGBM tests working_dir: lightgbm_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_aws.yaml run: timeout: 900 script: python workloads/ft_small_non_elastic.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_gce.yaml alert: default - name: lightgbm_tune_small group: LightGBM tests working_dir: lightgbm_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_aws.yaml run: timeout: 600 script: python workloads/tune_small.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_small_gce.yaml alert: default - name: lightgbm_tune_16x4 group: LightGBM tests working_dir: lightgbm_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_aws.yaml run: timeout: 900 script: python workloads/tune_16x4.py wait_for_nodes: num_nodes: 32 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_gce.yaml alert: default - name: lightgbm_tune_4x16 group: LightGBM tests working_dir: lightgbm_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_aws.yaml run: timeout: 900 script: python workloads/tune_4x16.py wait_for_nodes: num_nodes: 32 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_moderate_gce.yaml alert: default ####################### # Lightning tests ####################### # Naming convention: lightning_{accelerator}_{mode}_{#cpu}_{#gpu} - name: lightning_gpu_train_3x16_3x1 group: Lightning tests working_dir: lightning_tests frequency: nightly-3x team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_aws.yaml run: timeout: 1200 script: python workloads/test_trainer.py wait_for_nodes: num_nodes: 3 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gce.yaml alert: default - name: lightning_gpu_tune_3x16_3x1 group: Lightning tests working_dir: lightning_tests frequency: nightly-3x team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_aws.yaml run: timeout: 1200 script: python workloads/test_tuner.py wait_for_nodes: num_nodes: 3 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gce.yaml alert: default ####################### # ML user tests ####################### - name: ml_user_horovod_user_test_latest group: ML user tests working_dir: ml_user_tests frequency: nightly-3x team: ml cluster: cluster_env: horovod/app_config.yaml cluster_compute: horovod/compute_tpl_aws.yaml run: timeout: 1200 script: python horovod/horovod_user_test.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: horovod/app_config.yaml cluster_compute: horovod/compute_tpl_gce.yaml alert: default - name: ml_user_horovod_user_test_master group: ML user tests working_dir: ml_user_tests frequency: nightly-3x team: ml cluster: cluster_env: horovod/app_config_master.yaml cluster_compute: horovod/compute_tpl_aws.yaml run: timeout: 1200 script: python horovod/horovod_user_test.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: horovod/app_config_master.yaml cluster_compute: horovod/compute_tpl_gce.yaml alert: default - name: ml_user_train_tensorflow_mnist_test group: ML user tests working_dir: ml_user_tests frequency: nightly-3x team: ml cluster: cluster_env: train/app_config.yaml cluster_compute: train/compute_tpl_aws.yaml run: timeout: 36000 script: python train/train_tensorflow_mnist_test.py wait_for_nodes: num_nodes: 3 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: train/app_config.yaml cluster_compute: train/compute_tpl_gce.yaml alert: default - name: ml_user_train_torch_linear_test group: ML user tests working_dir: ml_user_tests frequency: nightly-3x team: ml cluster: cluster_env: train/app_config.yaml cluster_compute: train/compute_tpl_aws.yaml run: timeout: 36000 script: python train/train_torch_linear_test.py wait_for_nodes: num_nodes: 3 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: train/app_config.yaml cluster_compute: train/compute_tpl_gce.yaml alert: default - name: ml_user_xgboost_gpu_connect_latest group: ML user tests working_dir: ml_user_tests frequency: nightly-3x team: ml cluster: cluster_env: xgboost/app_config_gpu.yaml cluster_compute: xgboost/tpl_gpu_small_scaling_aws.yaml run: timeout: 1200 script: python xgboost/train_gpu_connect.py wait_for_nodes: num_nodes: 5 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: xgboost/app_config_gpu.yaml cluster_compute: xgboost/tpl_gpu_small_scaling_gce.yaml alert: default - name: ml_user_xgboost_gpu_connect_master group: ML user tests working_dir: ml_user_tests frequency: nightly-3x team: ml cluster: cluster_env: xgboost/app_config_gpu_master.yaml cluster_compute: xgboost/tpl_gpu_small_scaling_aws.yaml run: timeout: 1200 script: python xgboost/train_gpu_connect.py wait_for_nodes: num_nodes: 5 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: xgboost/app_config_gpu_master.yaml cluster_compute: xgboost/tpl_gpu_small_scaling_gce.yaml alert: default - name: ml_user_ray_lightning_user_test_latest group: ML user tests working_dir: ml_user_tests frequency: nightly-3x team: ml cluster: cluster_env: ray-lightning/app_config.yaml cluster_compute: ray-lightning/compute_tpl_aws.yaml run: timeout: 1200 script: python ray-lightning/ray_lightning_user_test.py wait_for_nodes: num_nodes: 3 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: ray-lightning/app_config.yaml cluster_compute: ray-lightning/compute_tpl_gce.yaml alert: default - name: ml_user_ray_lightning_user_test_master group: ML user tests working_dir: ml_user_tests frequency: nightly-3x team: ml cluster: cluster_env: ray-lightning/app_config_master.yaml cluster_compute: ray-lightning/compute_tpl_aws.yaml run: timeout: 1200 script: python ray-lightning/ray_lightning_user_test.py wait_for_nodes: num_nodes: 3 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: ray-lightning/app_config_master.yaml cluster_compute: ray-lightning/compute_tpl_gce.yaml alert: default - name: ml_user_tune_rllib_connect_test group: ML user tests working_dir: ml_user_tests frequency: nightly-3x team: ml cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: tune_rllib/compute_tpl_aws.yaml run: timeout: 2000 script: python tune_rllib/run_connect_tests.py wait_for_nodes: num_nodes: 9 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: tune_rllib/compute_tpl_gce.yaml alert: default ####################### # Tune cloud tests ####################### - name: tune_cloud_no_sync_down group: Tune cloud tests working_dir: tune_tests/cloud_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_aws_4x2.yaml run: timeout: 600 script: python workloads/run_cloud_test.py no_sync_down wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_compute: tpl_gce_4x8.yaml alert: tune_tests - name: tune_cloud_ssh_sync group: Tune cloud tests working_dir: tune_tests/cloud_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_aws_4x2.yaml run: timeout: 600 script: python workloads/run_cloud_test.py ssh_sync wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_gce_4x8.yaml alert: tune_tests - name: tune_cloud_durable_upload group: Tune cloud tests working_dir: tune_tests/cloud_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_aws_4x2.yaml run: timeout: 600 script: python workloads/run_cloud_test.py durable_upload --bucket s3://tune-cloud-tests/durable_upload wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_gce_4x8.yaml run: timeout: 600 script: python workloads/run_cloud_test.py durable_upload --bucket gs://tune-cloud-tests/durable_upload wait_for_nodes: num_nodes: 4 alert: tune_tests - name: tune_cloud_durable_upload_rllib_str group: Tune cloud tests working_dir: tune_tests/cloud_tests stable: false frequency: nightly team: ml cluster: cluster_env: app_config_ml.yaml cluster_compute: tpl_aws_4x2.yaml run: timeout: 600 script: python workloads/run_cloud_test.py durable_upload --trainable rllib_str --bucket s3://tune-cloud-tests/durable_upload_rllib_str wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config_ml.yaml cluster_compute: tpl_gce_4x2.yaml run: timeout: 600 script: python workloads/run_cloud_test.py durable_upload --trainable rllib_str --bucket gs://tune-cloud-tests/durable_upload_rllib_str wait_for_nodes: num_nodes: 4 alert: tune_tests - name: tune_cloud_durable_upload_rllib_trainer group: Tune cloud tests working_dir: tune_tests/cloud_tests stable: false frequency: nightly team: ml cluster: cluster_env: app_config_ml.yaml cluster_compute: tpl_aws_4x2.yaml run: timeout: 600 script: python workloads/run_cloud_test.py durable_upload --trainable rllib_trainer --bucket s3://tune-cloud-tests/durable_upload_rllib_trainer wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config_ml.yaml cluster_compute: tpl_gce_4x2.yaml run: timeout: 600 script: python workloads/run_cloud_test.py durable_upload --trainable rllib_str --bucket gs://tune-cloud-tests/durable_upload_rllib_trainer wait_for_nodes: num_nodes: 4 alert: tune_tests ######################## # Tune scalability tests ######################## - name: tune_scalability_bookkeeping_overhead group: Tune scalability tests working_dir: tune_tests/scalability_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_1x16.yaml run: timeout: 1200 script: python workloads/test_bookkeeping_overhead.py alert: tune_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_gce_1x16.yaml - name: tune_scalability_durable_trainable group: Tune scalability tests working_dir: tune_tests/scalability_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_16x2.yaml run: timeout: 900 script: python workloads/test_durable_trainable.py --bucket s3://tune-cloud-tests/scalability_durable_trainable wait_for_nodes: num_nodes: 16 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual run: timeout: 900 script: python workloads/test_durable_trainable.py --bucket gs://tune-cloud-tests/scalability_durable_trainable wait_for_nodes: num_nodes: 16 cluster: cluster_env: app_config.yaml cluster_compute: tpl_gce_16x2.yaml alert: tune_tests - name: tune_scalability_durable_multifile_checkpoints group: Tune scalability tests working_dir: tune_tests/scalability_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_16x2.yaml run: timeout: 900 script: python workloads/test_durable_multifile_checkpoints.py --bucket s3://tune-cloud-tests/scalability_durable_multifile_checkpoints wait_for_nodes: num_nodes: 16 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual run: timeout: 900 script: python workloads/test_durable_multifile_checkpoints.py --bucket gs://tune-cloud-tests/scalability_durable_multifile_checkpoints wait_for_nodes: num_nodes: 16 cluster: cluster_env: app_config.yaml cluster_compute: tpl_gce_16x2.yaml alert: tune_tests - name: tune_scalability_long_running_large_checkpoints group: Tune scalability tests working_dir: tune_tests/scalability_tests frequency: weekly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_1x32_hd.yaml run: timeout: 86400 script: python workloads/test_long_running_large_checkpoints.py long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: tune_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_gce_1x32_hd.yaml - name: tune_scalability_network_overhead group: Tune scalability tests working_dir: tune_tests/scalability_tests frequency: weekly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_100x2.yaml run: timeout: 900 prepare_timeout: 1200 script: python workloads/test_network_overhead.py wait_for_nodes: num_nodes: 100 alert: tune_tests variations: - __suffix__: aws - __suffix__: smoke-test frequency: nightly cluster: cluster_env: app_config.yaml cluster_compute: tpl_20x2.yaml run: timeout: 500 prepare_timeout: 600 script: python workloads/test_network_overhead.py --smoke-test wait_for_nodes: num_nodes: 20 - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_gce_100x2.yaml - name: tune_scalability_result_throughput_cluster group: Tune scalability tests working_dir: tune_tests/scalability_tests frequency: nightly-3x team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_16x64.yaml run: timeout: 600 script: python workloads/test_result_throughput_cluster.py wait_for_nodes: num_nodes: 16 alert: tune_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_gce_16x64.yaml - name: tune_scalability_result_throughput_single_node group: Tune scalability tests working_dir: tune_tests/scalability_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_1x96.yaml run: timeout: 600 script: python workloads/test_result_throughput_single_node.py alert: tune_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_gce_1x96.yaml run: timeout: 600 script: python workloads/test_result_throughput_single_node.py - name: tune_scalability_xgboost_sweep group: Tune scalability tests working_dir: tune_tests/scalability_tests frequency: weekly team: ml cluster: cluster_env: app_config_data.yaml cluster_compute: tpl_16x64.yaml run: timeout: 3600 script: python workloads/test_xgboost_sweep.py wait_for_nodes: num_nodes: 16 alert: tune_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config_data.yaml cluster_compute: tpl_gce_16x64.yaml ############################ # Tune fault tolerance tests ############################ - name: tune_worker_fault_tolerance group: Tune fault tolerance tests working_dir: tune_tests/fault_tolerance_tests stable: true frequency: nightly-3x team: ml cluster: cluster_env: app_config.yaml cluster_compute: tpl_aws_16x1.yaml run: timeout: 5400 script: python workloads/test_tune_worker_fault_tolerance.py --bucket s3://tune-cloud-tests/worker_fault_tolerance wait_for_nodes: num_nodes: 16 # Disabled until we can kill nodes in GCE # variations: # - __suffix__: aws # - __suffix__: gce # env: gce # frequency: manual # run: # timeout: 5400 # script: python workloads/test_tune_worker_fault_tolerance.py --bucket gs://tune-cloud-tests/worker_fault_tolerance # # wait_for_nodes: # num_nodes: 16 # cluster: # cluster_env: app_config.yaml # cluster_compute: tpl_gce_16x1.yaml ######################## # Golden Notebook tests ######################## - name: golden_notebook_torch_tune_serve_test group: Golden Notebook tests working_dir: golden_notebook_tests frequency: nightly-3x team: ml cluster: cluster_env: torch_tune_serve_app_config.yaml cluster_compute: gpu_tpl_aws.yaml run: timeout: 600 script: python workloads/torch_tune_serve_test.py wait_for_nodes: num_nodes: 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: torch_tune_serve_app_config.yaml cluster_compute: gpu_tpl_gce.yaml alert: default ####################### # Long running tests ####################### - name: long_running_actor_deaths group: Long running tests working_dir: long_running_tests frequency: weekly python: "3.8" team: core cluster: byod: runtime_env: - RLLIB_TEST_NO_JAX_IMPORT=1 cluster_env: app_config.yaml cluster_compute: tpl_cpu_1.yaml run: timeout: 86400 script: python workloads/actor_deaths.py long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_gce.yaml - name: long_running_apex group: Long running tests working_dir: long_running_tests frequency: weekly team: rllib cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: tpl_cpu_3.yaml run: timeout: 86400 script: python workloads/apex.py long_running: true wait_for_nodes: num_nodes: 3 smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: tpl_cpu_3_gce.yaml - name: long_running_impala group: Long running tests working_dir: long_running_tests frequency: weekly team: rllib cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: tpl_cpu_1_large.yaml run: timeout: 86400 script: python workloads/impala.py long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: tpl_cpu_1_large_gce.yaml - name: long_running_many_actor_tasks group: Long running tests working_dir: long_running_tests frequency: weekly python: "3.8" team: core cluster: byod: runtime_env: - RLLIB_TEST_NO_JAX_IMPORT=1 cluster_env: app_config.yaml cluster_compute: tpl_cpu_1.yaml run: timeout: 86400 script: python workloads/many_actor_tasks.py long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_gce.yaml - name: long_running_many_drivers group: Long running tests working_dir: long_running_tests frequency: weekly python: "3.8" team: core cluster: byod: runtime_env: - RLLIB_TEST_NO_JAX_IMPORT=1 cluster_env: app_config.yaml cluster_compute: tpl_cpu_1.yaml run: timeout: 86400 script: python workloads/many_drivers.py --iteration-num=4000 long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_gce.yaml - name: long_running_many_ppo group: Long running tests working_dir: long_running_tests stable: false frequency: weekly team: ml cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: many_ppo.yaml run: timeout: 86400 script: python workloads/many_ppo.py long_running: true wait_for_nodes: num_nodes: 1 smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: many_ppo_gce.yaml - name: long_running_many_tasks group: Long running tests working_dir: long_running_tests frequency: weekly python: "3.8" team: core cluster: byod: runtime_env: - RLLIB_TEST_NO_JAX_IMPORT=1 cluster_env: app_config.yaml cluster_compute: tpl_cpu_1.yaml run: timeout: 86400 script: python workloads/many_tasks.py long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_gce.yaml - name: long_running_many_tasks_serialized_ids group: Long running tests working_dir: long_running_tests frequency: weekly python: "3.8" team: core cluster: byod: runtime_env: - RLLIB_TEST_NO_JAX_IMPORT=1 cluster_env: app_config.yaml cluster_compute: tpl_cpu_1.yaml run: timeout: 86400 script: python workloads/many_tasks_serialized_ids.py long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_gce.yaml - name: long_running_node_failures group: Long running tests working_dir: long_running_tests frequency: weekly python: "3.8" team: core cluster: byod: runtime_env: - RLLIB_TEST_NO_JAX_IMPORT=1 cluster_env: app_config.yaml cluster_compute: tpl_cpu_1.yaml run: timeout: 86400 script: python workloads/node_failures.py long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_gce.yaml - name: long_running_pbt group: Long running tests working_dir: long_running_tests frequency: weekly team: ml cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: tpl_cpu_1.yaml run: timeout: 86400 script: python workloads/pbt.py long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: ../rllib_tests/app_config.yaml cluster_compute: tpl_cpu_1_gce.yaml - name: long_running_serve group: Long running tests working_dir: long_running_tests frequency: weekly team: serve cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1.yaml run: timeout: 86400 script: python workloads/serve.py long_running: true smoke_test: frequency: nightly run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_gce.yaml - name: long_running_serve_failure group: Long running tests working_dir: long_running_tests stable: true frequency: weekly team: serve cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_c5.yaml run: timeout: 86400 script: python workloads/serve_failure.py long_running: true smoke_test: frequency: nightly run: timeout: 600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 86400 cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_c5_gce.yaml - name: long_running_many_jobs group: Long running tests working_dir: long_running_tests stable: true frequency: weekly team: serve cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1.yaml run: timeout: 86400 script: python workloads/long_running_many_jobs.py --num-clients=1 long_running: true smoke_test: frequency: nightly run: timeout: 1800 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: app_config.yaml cluster_compute: tpl_cpu_1_gce.yaml - name: long_running_distributed_pytorch_pbt_failure group: Long running tests working_dir: long_running_distributed_tests frequency: weekly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl.yaml run: timeout: 86400 script: python workloads/pytorch_pbt_failure.py long_running: true smoke_test: frequency: manual run: timeout: 3600 alert: long_running_tests variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 3600 cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gce.yaml ######################## # Jobs tests ######################## - name: jobs_basic_local_working_dir group: Jobs tests working_dir: jobs_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_4_xlarge.yaml run: timeout: 600 script: python workloads/jobs_basic.py --working-dir "workloads" wait_for_nodes: num_nodes: 4 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gce_4_xlarge.yaml - name: jobs_basic_remote_working_dir group: Jobs tests working_dir: jobs_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_4_xlarge.yaml run: timeout: 600 script: python workloads/jobs_basic.py --working-dir "https://github.com/anyscale/job-services-cuj-examples/archive/refs/heads/main.zip" wait_for_nodes: num_nodes: 4 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gce_4_xlarge.yaml - name: jobs_remote_multi_node group: Jobs tests team: serve frequency: nightly working_dir: jobs_tests cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_4_xlarge.yaml run: timeout: 600 script: python workloads/jobs_remote_multi_node.py wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gce_4_xlarge.yaml - name: jobs_check_cuda_available group: Jobs tests team: serve frequency: nightly working_dir: jobs_tests cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gpu_node.yaml run: timeout: 600 script: python workloads/jobs_check_cuda_available.py wait_for_nodes: num_nodes: 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gce_gpu_node.yaml - name: jobs_specify_num_gpus group: Jobs tests team: serve frequency: nightly working_dir: jobs_tests cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gpu_worker.yaml run: timeout: 600 script: python workloads/jobs_specify_num_gpus.py --working-dir "workloads" wait_for_nodes: num_nodes: 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gce_gpu_worker.yaml ######################## # Runtime env tests ######################## - name: runtime_env_rte_many_tasks_actors group: Runtime env tests working_dir: runtime_env_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: rte_small.yaml run: timeout: 600 script: python workloads/rte_many_tasks_actors.py wait_for_nodes: num_nodes: 4 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: rte_gce_small.yaml - name: runtime_env_wheel_urls group: Runtime env tests working_dir: runtime_env_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: rte_minimal.yaml run: timeout: 9000 script: python workloads/wheel_urls.py wait_for_nodes: num_nodes: 1 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: rte_gce_minimal.yaml # It seems like the consensus is that this should be tested in CI, and not in a nightly test. # - name: runtime_env_rte_ray_client # group: Runtime env tests # working_dir: runtime_env_tests # frequency: nightly # team: serve # cluster: # cluster_env: app_config.yaml # cluster_compute: rte_minimal.yaml # run: # timeout: 600 # script: python workloads/rte_ray_client.py # wait_for_nodes: # num_nodes: 1 # alert: default ######################## # Serve tests ######################## - name: serve_single_deployment_1k_noop_replica group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_32_cpu.yaml run: timeout: 7200 long_running: false script: python workloads/single_deployment_1k_noop_replica.py alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_32_cpu_gce.yaml - name: serve_multi_deployment_1k_noop_replica group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_32_cpu.yaml run: timeout: 7200 long_running: false script: python workloads/multi_deployment_1k_noop_replica.py alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_32_cpu_gce.yaml - name: serve_autoscaling_single_deployment group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_8_cpu_autoscaling.yaml run: timeout: 7200 long_running: false script: python workloads/autoscaling_single_deployment.py alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_8_cpu_autoscaling_gce.yaml - name: serve_autoscaling_multi_deployment group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_32_cpu_autoscaling.yaml run: timeout: 7200 long_running: false script: python workloads/autoscaling_multi_deployment.py alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_32_cpu_autoscaling_gce.yaml - name: serve_serve_micro_benchmark group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node.yaml run: timeout: 7200 long_running: false script: python workloads/serve_micro_benchmark.py alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_gce.yaml # - name: serve_serve_micro_benchmark_k8s # group: Serve tests # working_dir: serve_tests # # TODO(architkulkarni) Reenable after K8s migration. Currently failing # frequency: manual # team: serve # cluster: # cluster_env: app_config.yaml # cluster_compute: compute_tpl_single_node_k8s.yaml # run: # timeout: 7200 # long_running: false # script: python workloads/serve_micro_benchmark.py # alert: default - name: deployment_graph_long_chain group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_32_cpu.yaml run: timeout: 3600 long_running: false script: python workloads/deployment_graph_long_chain.py --chain-length=10 --num-clients=4 --local-test=False alert: default stable: False variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_32_cpu_gce.yaml - name: deployment_graph_wide_ensemble group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_32_cpu.yaml run: timeout: 3600 long_running: false script: python workloads/deployment_graph_wide_ensemble.py --fanout-degree=10 --num-clients=4 --local-test=False alert: default stable: False variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_32_cpu_gce.yaml - name: serve_handle_long_chain group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_32_cpu.yaml run: timeout: 3600 long_running: false script: python workloads/serve_handle_long_chain.py --chain-length=10 --num-clients=4 --local-test=False alert: default stable: False variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_32_cpu_gce.yaml - name: serve_handle_wide_ensemble group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_32_cpu.yaml run: timeout: 3600 long_running: false script: python workloads/serve_handle_wide_ensemble.py --fanout-degree=10 --num-clients=4 --local-test=False alert: default stable: False variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_32_cpu_gce.yaml - name: serve_micro_protocol_grpc_benchmark group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node.yaml run: timeout: 7200 long_running: false script: python workloads/serve_protocol_benchmark.py --data-size=1048576 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_gce.yaml - name: serve_micro_protocol_http_benchmark group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node.yaml run: timeout: 7200 long_running: false script: python workloads/serve_protocol_benchmark.py --data-size=1048576 --http-test alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_single_node_gce.yaml - name: serve_resnet_benchmark group: Serve tests working_dir: serve_tests frequency: nightly team: serve cluster: cluster_env: gpu_app_config.yaml cluster_compute: compute_tpl_gpu_node.yaml run: timeout: 7200 long_running: false script: python workloads/serve_resnet_benchmark.py --gpu-env alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: gpu_app_config.yaml cluster_compute: compute_tpl_gpu_node_gce.yaml ######################## # Train tests ######################## - name: train_horovod_multi_node_test group: Train tests working_dir: train_tests/horovod frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_aws.yaml run: timeout: 3000 script: python train_horovod_multi_node_test.py wait_for_nodes: num_nodes: 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl_gce.yaml alert: default ######################## # Alpa tests ######################## - name: alpa_opt_2_7b_sanity_check group: Alpa tests working_dir: alpa_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: gpu_2x4_t4_aws.yaml run: timeout: 3600 script: bash run_train_opt_2_7b.sh --storage aws wait_for_nodes: num_nodes: 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: gpu_2x4_t4_gce.yaml run: timeout: 3600 script: bash run_train_opt_2_7b.sh --storage gcs wait_for_nodes: num_nodes: 2 alert: default - name: alpa_opt_30b_inference group: Alpa tests working_dir: alpa_tests frequency: nightly team: ml cluster: cluster_env: app_config.yaml cluster_compute: gpu_1x8_v100_aws.yaml run: timeout: 3600 script: bash run_inference_opt_30b.sh --storage aws wait_for_nodes: num_nodes: 1 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: gpu_1x8_v100_gce.yaml run: timeout: 3600 script: bash run_inference_opt_30b.sh --storage gcs wait_for_nodes: num_nodes: 1 alert: default ######################## # RLlib tests ######################## - name: rllib_learner_group_checkpointing_multinode group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: multi_node_checkpointing_compute_config.yaml run: timeout: 3600 script: pytest checkpointing_tests/test_learner_group_checkpointing.py wait_for_nodes: num_nodes: 3 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: multi_node_checkpointing_compute_config_gce.yaml - name: rllib_learner_e2e_module_loading group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: multi_node_checkpointing_compute_config.yaml run: timeout: 3600 script: pytest checkpointing_tests/test_e2e_rl_module_restore.py wait_for_nodes: num_nodes: 3 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: multi_node_checkpointing_compute_config_gce.yaml - name: rllib_multi_node_e2e_training_smoke_test group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: multi_node_checkpointing_compute_config.yaml run: timeout: 3600 script: pytest smoke_tests/smoke_test_basic_multi_node_training_learner.py wait_for_nodes: num_nodes: 3 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: multi_node_checkpointing_compute_config_gce.yaml - name: rllib_learning_tests_a2c_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=a2c --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_a2c_torch group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=a2c --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_a3c_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 32cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=a3c --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 32cpus_gce.yaml - name: rllib_learning_tests_apex_tf group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_24cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=apex --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_24cpus_gce.yaml - name: rllib_learning_tests_apex_torch group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_24cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=apex --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_24cpus_gce.yaml - name: rllib_learning_tests_appo_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_64cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=appo/new_stack --framework=tf2 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_64cpus_gce.yaml - name: rllib_learning_tests_appo_torch group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_64cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=appo/new_stack --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_64cpus_gce.yaml # TODO (sven): Remove this test once we are on new stack by default for APPO. - name: rllib_learning_tests_appo_old_stack_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_32cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=appo/old_stack --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_32cpus_gce.yaml # TODO (sven): Remove this test once we are on new stack by default for APPO. - name: rllib_learning_tests_appo_old_stack_torch group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_32cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=appo/old_stack --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_32cpus_gce.yaml - name: rllib_learning_tests_bc_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=bc --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_bc_torch group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=bc --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_cql_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib # Marking as unstable since it's currently expected to fail. stable: false cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=cql --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_cql_torch group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=cql --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_ddpg_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=ddpg --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_ddpg_torch group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=ddpg --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_dqn_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=dqn --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_dqn_torch group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=dqn --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_es_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_64cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=es --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_64cpus_gce.yaml - name: rllib_learning_tests_es_torch group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_64cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=es --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_64cpus_gce.yaml - name: rllib_learning_tests_impala_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=impala --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_impala_torch group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=impala --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_marwil_tf group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=marwil --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_marwil_torch group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=marwil --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_ppo_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_64cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=ppo/new_stack --framework=tf2 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_64cpus_gce.yaml - name: rllib_learning_tests_ppo_torch group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_64cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=ppo/new_stack --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_64cpus_gce.yaml # TODO (sven): Remove this test once we are on new stack by default for APPO. - name: rllib_learning_tests_ppo_old_stack_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_32cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=ppo/old_stack --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_32cpus_gce.yaml # TODO (sven): Remove this test once we are on new stack by default for APPO. - name: rllib_learning_tests_ppo_old_stack_torch group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_32cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=ppo/old_stack --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 2gpus_32cpus_gce.yaml - name: rllib_learning_tests_sac_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=sac --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_sac_torch group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=sac --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_slateq_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=slateq --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_slateq_torch group: RLlib tests working_dir: rllib_tests # Marking as unstable since it's currently expected to fail. stable: false frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=slateq --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_td3_tf group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=td3 --framework=tf alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_learning_tests_td3_torch group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus.yaml run: timeout: 18000 script: python learning_tests/run.py --yaml-sub-dir=td3 --framework=torch alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 1gpu_16cpus_gce.yaml - name: rllib_multi_gpu_learning_tests group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 8gpus_96cpus.yaml run: timeout: 7200 script: python multi_gpu_learning_tests/run.py alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 8gpus_96cpus_gce.yaml - name: rllib_multi_gpu_with_lstm_learning_tests group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 8gpus_96cpus.yaml run: timeout: 7200 script: python multi_gpu_with_lstm_learning_tests/run.py alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: 8gpus_96cpus_gce.yaml - name: rllib_multi_gpu_with_attention_learning_tests group: RLlib tests working_dir: rllib_tests frequency: nightly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 8gpus_96cpus.yaml run: timeout: 7200 script: python multi_gpu_with_attention_learning_tests/run.py alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: # TODO(https://github.com/ray-project/ray/issues/34591) # Revert to the comment below once ^ closed. # cluster_env: app_config.yaml cluster_env: debug_app_config.yaml cluster_compute: 8gpus_96cpus_gce.yaml - name: rllib_stress_tests group: RLlib tests working_dir: rllib_tests frequency: weekly team: rllib cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_544_cpus.yaml run: timeout: 5400 script: python stress_tests/run_stress_tests.py wait_for_nodes: num_nodes: 6 smoke_test: frequency: nightly run: timeout: 2000 alert: default variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual smoke_test: frequency: manual run: timeout: 2000 cluster: cluster_env: app_config.yaml cluster_compute: 4gpus_512_cpus_gce.yaml ######################## # Core Nightly Tests ######################## - name: shuffle_100gb group: core-multi-test working_dir: nightly_tests frequency: nightly python: "3.8" team: core cluster: byod: runtime_env: - RAY_worker_killing_policy=retriable_lifo cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/shuffle_compute_multi.yaml run: timeout: 3000 script: python shuffle/shuffle_test.py --num-partitions=200 --partition-size=500e6 wait_for_nodes: num_nodes: 4 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/shuffle_compute_multi_gce.yaml - name: stress_test_placement_group group: core-multi-test working_dir: nightly_tests frequency: nightly python: "3.8" team: core cluster: byod: {} cluster_env: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/placement_group_tests_compute.yaml run: timeout: 7200 script: python stress_tests/test_placement_group.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/placement_group_tests_compute_gce.yaml - name: decision_tree_autoscaling_20_runs group: core-multi-test working_dir: nightly_tests frequency: nightly python: "3.8" team: core cluster: byod: {} cluster_env: decision_tree/decision_tree_app_config.yaml cluster_compute: decision_tree/autoscaling_compute.yaml run: timeout: 9600 script: python decision_tree/cart_with_tree.py --concurrency=20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: decision_tree/decision_tree_app_config.yaml cluster_compute: decision_tree/autoscaling_compute_gce.yaml - name: autoscaling_shuffle_1tb_1000_partitions group: core-multi-test working_dir: nightly_tests frequency: nightly python: "3.8" team: core cluster: byod: runtime_env: - RAY_worker_killing_policy=retriable_lifo cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/shuffle_compute_autoscaling.yaml run: timeout: 4000 script: python shuffle/shuffle_test.py --num-partitions=1000 --partition-size=1e9 --no-streaming variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/shuffle_compute_autoscaling_gce.yaml - name: microbenchmark group: core-daily-test team: core frequency: nightly working_dir: microbenchmark python: "3.7" cluster: cluster_env: app_config.yaml cluster_compute: tpl_64.yaml run: timeout: 1800 script: OMP_NUM_THREADS=64 RAY_ADDRESS=local python run_microbenchmark.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_64_gce.yaml - name: microbenchmark_38 group: core-daily-test team: core frequency: nightly working_dir: microbenchmark python: "3.8" cluster: byod: {} cluster_env: app_config.yaml cluster_compute: tpl_64.yaml run: timeout: 1800 script: OMP_NUM_THREADS=64 RAY_ADDRESS=local python run_microbenchmark.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: tpl_64_gce.yaml - name: benchmark_worker_startup group: core-daily-test team: core frequency: nightly working_dir: benchmark-worker-startup stable: false python: "3.9" cluster: cluster_env: app_config_gpu.yaml cluster_compute: only_head_node_1gpu_64cpu.yaml run: timeout: 7200 script: python benchmark_worker_startup.py --num_cpus_in_cluster 64 --num_gpus_in_cluster 64 --num_tasks_or_actors_per_run 64 --num_measurements_per_configuration 5 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config_gpu.yaml cluster_compute: only_head_node_1gpu_64cpu_gce.yaml - name: dask_on_ray_100gb_sort group: core-daily-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: dask_on_ray/dask_on_ray_app_config.yaml cluster_compute: dask_on_ray/dask_on_ray_sort_compute_template.yaml run: timeout: 7200 script: python dask_on_ray/dask_on_ray_sort.py --nbytes 100_000_000_000 --npartitions 200 --num-nodes 1 --ray --data-dir /tmp/ray --file-path /tmp/ray variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: dask_on_ray/dask_on_ray_app_config.yaml cluster_compute: dask_on_ray/dask_on_ray_sort_compute_template_gce.yaml - name: dask_on_ray_large_scale_test_spilling group: core-daily-test working_dir: nightly_tests frequency: nightly team: data cluster: cluster_env: dask_on_ray/large_scale_dask_on_ray_app_config.yaml cluster_compute: dask_on_ray/dask_on_ray_stress_compute.yaml run: timeout: 7200 script: python dask_on_ray/large_scale_test.py --num_workers 150 --worker_obj_store_size_in_gb 70 --error_rate 0 --data_save_path /tmp/ray wait_for_nodes: num_nodes: 21 smoke_test: frequency: nightly cluster: app_config: dask_on_ray/large_scale_dask_on_ray_app_config.yaml cluster_compute: dask_on_ray/large_scale_dask_on_ray_compute_template.yaml run: timeout: 7200 script: python dask_on_ray/large_scale_test.py --num_workers 32 --worker_obj_store_size_in_gb 70 --error_rate 0 --data_save_path /tmp/ray wait_for_nodes: num_nodes: 5 - name: stress_test_state_api_scale group: core-daily-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: stress_tests/state_api_app_config.yaml cluster_compute: stress_tests/stress_tests_compute_large.yaml run: timeout: 3600 script: python stress_tests/test_state_api_scale.py wait_for_nodes: num_nodes: 7 smoke_test: frequency: nightly cluster: app_config: stress_tests/state_api_app_config.yaml cluster_compute: stress_tests/smoke_test_compute.yaml run: timeout: 3600 wait_for_nodes: num_nodes: 5 script: python stress_tests/test_state_api_scale.py --smoke-test variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: stress_tests/state_api_app_config.yaml cluster_compute: stress_tests/stress_tests_compute_large_gce.yaml smoke_test: frequency: manual - name: shuffle_20gb_with_state_api group: core-daily-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: shuffle/shuffle_with_state_api_app_config.yaml cluster_compute: shuffle/shuffle_compute_single.yaml run: timeout: 1000 script: python stress_tests/test_state_api_with_other_tests.py nightly_tests/shuffle/shuffle_test.py --test-args="--num-partitions=100 --partition-size=200e6" variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_with_state_api_app_config.yaml cluster_compute: shuffle/shuffle_compute_single_gce.yaml - name: stress_test_many_tasks group: core-daily-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/stress_tests_compute.yaml run: timeout: 14400 wait_for_nodes: num_nodes: 101 script: python stress_tests/test_many_tasks.py smoke_test: frequency: nightly cluster: app_config: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/smoke_test_compute.yaml run: timeout: 3600 wait_for_nodes: num_nodes: 5 script: python stress_tests/test_many_tasks.py --num-nodes=4 --smoke-test variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/stress_tests_compute_gce.yaml smoke_test: frequency: manual - name: stress_test_dead_actors group: core-daily-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/stress_tests_compute.yaml run: timeout: 7200 wait_for_nodes: num_nodes: 101 script: python stress_tests/test_dead_actors.py smoke_test: frequency: nightly cluster: app_config: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/smoke_test_compute.yaml run: timeout: 3600 wait_for_nodes: num_nodes: 5 script: python stress_tests/test_dead_actors.py --num-nodes=4 --num-parents=3 --num-children=3 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/stress_tests_compute_gce.yaml smoke_test: frequency: manual # The full test is not stable, so run the smoke test only. # See https://github.com/ray-project/ray/issues/23244. - name: threaded_actors_stress_test group: core-daily-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/smoke_test_compute.yaml run: timeout: 3600 script: python stress_tests/test_threaded_actors.py --test-runtime 1800 --kill-interval_s 30 wait_for_nodes: num_nodes: 5 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: stress_tests/stress_tests_app_config.yaml cluster_compute: stress_tests/smoke_test_compute_gce.yaml # - name: threaded_actors_stress_test # group: core-daily-test # working_dir: nightly_tests # # frequency: nightly # team: core # cluster: # cluster_env: stress_tests/stress_tests_app_config.yaml # cluster_compute: stress_tests/stress_test_threaded_actor_compute.yaml # # run: # timeout: 7200 # script: python stress_tests/test_threaded_actors.py --test-runtime 3600 --kill-interval_s # 60 # # wait_for_nodes: # num_nodes: 201 # timeout: 600 # # smoke_test: # frequency: nightly # cluster: # app_config: stress_tests/stress_tests_app_config.yaml # cluster_compute: stress_tests/smoke_test_compute.yaml # # run: # timeout: 3600 # script: python stress_tests/test_threaded_actors.py --test-runtime 1800 --kill-interval_s # 30 # # wait_for_nodes: # num_nodes: 5 # timeout: 600 - name: single_node_oom group: core-daily-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: stress_tests/stress_tests_single_node_oom_app_config.yaml cluster_compute: stress_tests/stress_tests_single_node_oom_compute.yaml run: timeout: 500 script: python stress_tests/test_parallel_tasks_memory_pressure.py --num-tasks 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: stress_tests/stress_tests_single_node_oom_app_config.yaml cluster_compute: stress_tests/stress_tests_single_node_oom_compute_gce.yaml - name: tune_air_oom group: core-daily-test working_dir: air_tests stable: false jailed: true frequency: nightly team: core cluster: cluster_env: oom/stress_tests_tune_air_oom_app_config.yaml cluster_compute: oom/stress_tests_tune_air_oom_compute.yaml run: timeout: 3600 script: bash oom/tune_air_oom.sh - name: dask_on_ray_1tb_sort group: core-daily-test working_dir: nightly_tests frequency: nightly-3x team: core cluster: cluster_env: dask_on_ray/dask_on_ray_app_config.yaml cluster_compute: dask_on_ray/1tb_sort_compute.yaml run: timeout: 7200 script: python dask_on_ray/dask_on_ray_sort.py --nbytes 1_000_000_000_000 --npartitions 1000 --num-nodes 31 --ray --data-dir /tmp/ray --s3-bucket core-nightly-test wait_for_nodes: num_nodes: 32 - name: many_nodes_actor_test_on_v2 group: core-daily-test working_dir: benchmarks frequency: nightly-3x team: core cluster: cluster_env: distributed/many_nodes_tests/app_config.yaml cluster_compute: distributed/many_nodes_tests/compute_config.yaml run: timeout: 3600 # 2cpus per node x 1000 nodes / 0.2 cpus per actor = 10k # 2cpus per node x 2000 nodes / 0.2 cpus per actor = 20k script: python distributed/many_nodes_tests/actor_test.py --no-wait --cpus-per-actor=0.2 --total-actors 10000 20000 wait_for_nodes: num_nodes: 500 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: distributed/many_nodes_tests/app_config.yaml cluster_compute: distributed/many_nodes_tests/compute_config_gce.yaml #- name: many_nodes_multi_master_test # group: core-daily-test # working_dir: nightly_tests # # frequency: nightly-3x # team: core # cluster: # cluster_env: many_nodes_tests/app_config.yaml # cluster_compute: many_nodes_tests/compute_config.yaml # # run: # timeout: 7200 # script: python many_nodes_tests/multi_master_test.py # wait_for_nodes: # num_nodes: 251 # - name: pg_autoscaling_regression_test group: core-daily-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: placement_group_tests/app_config.yaml cluster_compute: placement_group_tests/compute.yaml run: timeout: 1200 script: python placement_group_tests/pg_run.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: placement_group_tests/app_config.yaml cluster_compute: placement_group_tests/compute_gce.yaml - name: placement_group_performance_test group: core-daily-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: placement_group_tests/app_config.yaml cluster_compute: placement_group_tests/pg_perf_test_compute.yaml run: timeout: 1200 script: python placement_group_tests/placement_group_performance_test.py wait_for_nodes: num_nodes: 5 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: placement_group_tests/app_config.yaml cluster_compute: placement_group_tests/pg_perf_test_compute_gce.yaml ######################### # Core Scalability Tests ######################### - name: single_node group: core-scalability-test working_dir: benchmarks frequency: nightly team: core cluster: cluster_env: app_config.yaml cluster_compute: single_node.yaml run: timeout: 12000 prepare: sleep 0 script: python single_node/test_single_node.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: single_node_gce.yaml - name: object_store group: core-scalability-test working_dir: benchmarks frequency: nightly team: core cluster: cluster_env: app_config.yaml cluster_compute: object_store.yaml run: timeout: 3600 script: python object_store/test_object_store.py wait_for_nodes: num_nodes: 50 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: object_store_gce.yaml - name: many_actors group: core-scalability-test working_dir: benchmarks frequency: nightly-3x team: core cluster: cluster_env: app_config.yaml cluster_compute: distributed.yaml run: timeout: 3600 script: python distributed/test_many_actors.py wait_for_nodes: num_nodes: 65 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: distributed_gce.yaml - name: many_actors_smoke_test group: core-scalability-test working_dir: benchmarks frequency: nightly team: core cluster: cluster_env: app_config.yaml cluster_compute: distributed_smoke_test.yaml run: timeout: 3600 script: SMOKE_TEST=1 python distributed/test_many_actors.py wait_for_nodes: num_nodes: 2 - name: many_tasks group: core-scalability-test working_dir: benchmarks frequency: nightly team: core cluster: cluster_env: app_config.yaml cluster_compute: distributed.yaml run: timeout: 3600 script: python distributed/test_many_tasks.py --num-tasks=10000 wait_for_nodes: num_nodes: 65 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: distributed_gce.yaml - name: many_pgs group: core-scalability-test working_dir: benchmarks frequency: nightly-3x team: core cluster: cluster_env: app_config.yaml cluster_compute: distributed.yaml run: timeout: 3600 script: python distributed/test_many_pgs.py wait_for_nodes: num_nodes: 65 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: distributed_gce.yaml - name: many_pgs_smoke_test group: core-scalability-test working_dir: benchmarks frequency: nightly team: core cluster: cluster_env: app_config.yaml cluster_compute: distributed_smoke_test.yaml run: timeout: 3600 script: SMOKE_TEST=1 python distributed/test_many_pgs.py wait_for_nodes: num_nodes: 2 - name: many_nodes group: core-scalability-test working_dir: benchmarks frequency: nightly-3x team: core cluster: cluster_env: app_config.yaml cluster_compute: many_nodes.yaml run: timeout: 3600 script: python distributed/test_many_tasks.py --num-tasks=1000 wait_for_nodes: num_nodes: 250 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: many_nodes_gce.yaml - name: scheduling_test_many_0s_tasks_many_nodes group: core-scalability-test working_dir: benchmarks frequency: nightly team: core cluster: cluster_env: app_config.yaml cluster_compute: scheduling.yaml run: timeout: 3600 script: python distributed/test_scheduling.py --total-num-task=1984000 --num-cpu-per-task=1 --task-duration-s=0 --total-num-actors=32 --num-actors-per-nodes=1 wait_for_nodes: num_nodes: 32 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: scheduling_gce.yaml # - name: scheduling_test_many_5s_tasks_single_node # group: core-scalability-test # working_dir: benchmarks # frequency: nightly # team: core # cluster: # cluster_env: app_config.yaml # cluster_compute: scheduling.yaml # run: # timeout: 3600 # script: python distributed/test_scheduling.py --total-num-task=1984000 --num-cpu-per-task=1 # --task-duration-s=5 --total-num-actors=1 --num-actors-per-nodes=1 # wait_for_nodes: # num_nodes: 32 # timeout: 600 # stable: false # - name: scheduling_test_many_5s_tasks_many_nodes # group: core-scalability-test # working_dir: benchmarks # frequency: nightly # team: core # cluster: # cluster_env: app_config.yaml # cluster_compute: scheduling.yaml # run: # timeout: 3600 # script: python distributed/test_scheduling.py --total-num-task=1984000 --num-cpu-per-task=1 # --task-duration-s=5 --total-num-actors=32 --num-actors-per-nodes=1 # wait_for_nodes: # num_nodes: 32 # timeout: 600 # stable: false ############### # Dataset tests ############### - name: inference group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: inference.yaml run: timeout: 600 script: python inference.py wait_for_nodes: num_nodes: 2 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: inference_gce.yaml - name: shuffle_data_loader group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: shuffle_app_config.yaml cluster_compute: shuffle_compute.yaml run: timeout: 1800 script: python dataset_shuffle_data_loader.py --cloud aws variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_compute: shuffle_compute_gce.yaml run: script: python dataset_shuffle_data_loader.py --cloud gcp - name: parquet_metadata_resolution group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute.yaml run: # Expect the test to finish around 40 seconds. timeout: 100 script: python parquet_metadata_resolution.py --num-files 915 --cloud aws variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_compute: single_node_benchmark_compute_gce.yaml run: script: python parquet_metadata_resolution.py --num-files 915 --cloud gcp - name: dataset_random_access group: data-tests working_dir: nightly_tests/dataset stable: false frequency: nightly team: data cluster: cluster_env: pipelined_training_app.yaml cluster_compute: pipelined_training_compute.yaml run: timeout: 1200 script: python dataset_random_access.py wait_for_nodes: num_nodes: 15 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: pipelined_training_app.yaml cluster_compute: pipelined_training_compute_gce.yaml - name: pipelined_data_ingest_benchmark_1tb group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: data_ingest_benchmark_compute.yaml run: timeout: 300 script: python data_ingest_benchmark.py --dataset-size-gb=1000 --num-workers=20 --streaming wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: data_ingest_benchmark_compute_gce.yaml - name: streaming_data_ingest_benchmark_1tb group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: data_ingest_benchmark_compute.yaml run: timeout: 300 script: python data_ingest_benchmark.py --dataset-size-gb=1000 --num-workers=20 --new_streaming wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: data_ingest_benchmark_compute_gce.yaml - name: streaming_data_ingest_benchmark_100gb_gpu group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: data_ingest_benchmark_compute_gpu.yaml run: timeout: 300 script: python data_ingest_benchmark.py --dataset-size-gb=100 --num-workers=4 --new_streaming --use-gpu wait_for_nodes: num_nodes: 3 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: data_ingest_benchmark_compute_gpu_gce.yaml # This test case will early stop the data ingestion iteration on the GPU actors. # This is a common usage in PyTorch Lightning # (https://lightning.ai/docs/pytorch/stable/common/trainer.html#limit-train-batches). # There was a bug in Ray Data that caused GPU memoy leak (see #34819). # We add this test case to cover this scenario. - name: streaming_data_ingest_benchmark_100gb_gpu_early_stop group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: data_ingest_benchmark_compute_gpu.yaml run: timeout: 300 script: python data_ingest_benchmark.py --dataset-size-gb=100 --num-workers=4 --new_streaming --use-gpu --early-stop wait_for_nodes: num_nodes: 3 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: data_ingest_benchmark_compute_gpu_gce.yaml - name: aggregate_benchmark group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute.yaml run: timeout: 1800 script: python aggregate_benchmark.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute_gce.yaml - name: read_parquet_benchmark_single_node group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute.yaml run: # Expect the benchmark to finish in 400 seconds. timeout: 400 script: python read_parquet_benchmark.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute_gce.yaml - name: read_images_benchmark_single_node group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute.yaml run: timeout: 1800 script: python read_images_benchmark.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute_gce.yaml - name: read_tfrecords_benchmark_single_node group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: read_tfrecords_benchmark_app.yaml cluster_compute: single_node_benchmark_compute.yaml run: # Expect the benchmark to finish around 22 minutes. timeout: 1800 script: python read_tfrecords_benchmark.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: read_tfrecords_benchmark_app.yaml cluster_compute: single_node_benchmark_compute_gce.yaml - name: map_batches_benchmark_single_node group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute.yaml run: # Expect the benchmark to finish around 30 minutes. timeout: 2400 script: python map_batches_benchmark.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute_gce.yaml - name: iter_tensor_batches_benchmark_single_node group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute.yaml run: # Expect the benchmark to finish around 30 minutes. timeout: 2400 script: python iter_tensor_batches_benchmark.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute_gce.yaml - name: iter_tensor_batches_benchmark_multi_node group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: multi_node_benchmark_compute.yaml run: # Expect the benchmark to finish around 30 minutes. timeout: 2400 script: python iter_tensor_batches_benchmark.py --data-size-gb=10 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: multi_node_benchmark_compute_gce.yaml - name: iter_batches_benchmark_single_node group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute.yaml run: # Expect the benchmark to finish around 12 minutes. timeout: 1080 script: python iter_batches_benchmark.py variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: app_config.yaml cluster_compute: single_node_benchmark_compute_gce.yaml - name: pipelined_training_50_gb group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: pipelined_training_app.yaml cluster_compute: pipelined_training_compute.yaml run: timeout: 4800 script: python pipelined_training.py --epochs 1 wait_for_nodes: num_nodes: 15 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: pipelined_training_app.yaml cluster_compute: pipelined_training_compute_gce.yaml - name: pipelined_ingestion_1500_gb group: data-tests working_dir: nightly_tests/dataset frequency: nightly team: data cluster: cluster_env: pipelined_ingestion_app.yaml cluster_compute: pipelined_ingestion_compute.yaml run: timeout: 9600 script: python pipelined_training.py --epochs 2 --num-windows 5 --num-files 915 --debug wait_for_nodes: num_nodes: 21 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: pipelined_training_app.yaml cluster_compute: pipelined_training_compute_gce.yaml - name: dataset_shuffle_random_shuffle_1tb group: data-tests working_dir: nightly_tests frequency: nightly team: data cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances.yaml run: timeout: 7200 script: python dataset/sort.py --num-partitions=1000 --partition-size=1e9 --shuffle wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances_gce.yaml - name: dataset_shuffle_sort_1tb group: data-tests working_dir: nightly_tests stable: false frequency: nightly team: data cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances.yaml run: timeout: 7200 script: python dataset/sort.py --num-partitions=1000 --partition-size=1e9 wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances_gce.yaml - name: dataset_shuffle_push_based_random_shuffle_1tb group: data-tests working_dir: nightly_tests frequency: nightly team: data cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances.yaml run: timeout: 7200 script: RAY_DATA_PUSH_BASED_SHUFFLE=1 python dataset/sort.py --num-partitions=1000 --partition-size=1e9 --shuffle wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances_gce.yaml - name: dataset_shuffle_push_based_sort_1tb group: data-tests working_dir: nightly_tests frequency: nightly team: data cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances.yaml run: timeout: 7200 script: RAY_DATA_PUSH_BASED_SHUFFLE=1 python dataset/sort.py --num-partitions=1000 --partition-size=1e9 wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances_gce.yaml - name: dataset_shuffle_push_based_random_shuffle_100tb group: data-tests working_dir: nightly_tests frequency: weekly team: data cluster: cluster_env: shuffle/100tb_shuffle_app_config.yaml cluster_compute: shuffle/100tb_shuffle_compute.yaml run: timeout: 28800 script: RAY_DATA_PUSH_BASED_SHUFFLE=1 python dataset/sort.py --num-partitions=100000 --partition-size=1e9 --shuffle wait_for_nodes: num_nodes: 100 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/100tb_shuffle_app_config_gce.yaml cluster_compute: shuffle/100tb_shuffle_compute_gce.yaml run: timeout: 28800 script: RAY_DATA_PUSH_BASED_SHUFFLE=1 python dataset/sort.py --num-partitions=40000 --partition-size=1e9 --shuffle wait_for_nodes: num_nodes: 100 ################## # Core Chaos tests ################## - name: chaos_many_tasks_no_object_store group: core-nightly-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: chaos_test/app_config.yaml cluster_compute: chaos_test/compute_template.yaml run: timeout: 3600 wait_for_nodes: num_nodes: 10 prepare: python setup_chaos.py --no-start script: python chaos_test/test_chaos_basic.py --workload=tasks variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: chaos_test/app_config.yaml cluster_compute: chaos_test/compute_template_gce.yaml - name: chaos_many_actors group: core-nightly-test working_dir: nightly_tests frequency: nightly team: core cluster: cluster_env: chaos_test/app_config.yaml cluster_compute: chaos_test/compute_template.yaml run: timeout: 4200 wait_for_nodes: num_nodes: 10 prepare: python setup_chaos.py --no-start script: python chaos_test/test_chaos_basic.py --workload=actors variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: chaos_test/app_config.yaml cluster_compute: chaos_test/compute_template_gce.yaml - name: chaos_dask_on_ray_large_scale_test_no_spilling group: data-tests working_dir: nightly_tests frequency: nightly team: data cluster: cluster_env: chaos_test/dask_on_ray_app_config_reconstruction.yaml cluster_compute: dask_on_ray/chaos_dask_on_ray_stress_compute.yaml run: timeout: 7200 wait_for_nodes: num_nodes: 21 prepare: python setup_chaos.py --node-kill-interval 100 script: python dask_on_ray/large_scale_test.py --num_workers 20 --worker_obj_store_size_in_gb 20 --error_rate 0 --data_save_path /tmp/ray variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: chaos_test/dask_on_ray_app_config_reconstruction.yaml cluster_compute: dask_on_ray/dask_on_ray_stress_compute_gce.yaml - name: chaos_dask_on_ray_large_scale_test_spilling group: data-tests working_dir: nightly_tests frequency: nightly team: data cluster: cluster_env: chaos_test/dask_on_ray_app_config_reconstruction.yaml cluster_compute: dask_on_ray/dask_on_ray_stress_compute.yaml run: timeout: 7200 wait_for_nodes: num_nodes: 21 prepare: python setup_chaos.py --node-kill-interval 100 script: python dask_on_ray/large_scale_test.py --num_workers 150 --worker_obj_store_size_in_gb 70 --error_rate 0 --data_save_path /tmp/ray variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: chaos_test/dask_on_ray_app_config_reconstruction.yaml cluster_compute: dask_on_ray/dask_on_ray_stress_compute_gce.yaml - name: chaos_pipelined_ingestion_1500_gb_15_windows group: data-tests working_dir: nightly_tests stable: false frequency: nightly team: data cluster: cluster_env: dataset/pipelined_ingestion_app.yaml cluster_compute: dataset/pipelined_ingestion_compute.yaml run: timeout: 7200 wait_for_nodes: num_nodes: 21 prepare: ' python setup_chaos.py --node-kill-interval 300' script: python dataset/pipelined_training.py --epochs 1 --num-windows 15 --num-files 915 --debug variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: dataset/pipelined_ingestion_app.yaml cluster_compute: dataset/pipelined_ingestion_compute_gce.yaml - name: chaos_dataset_shuffle_push_based_sort_1tb group: data-tests working_dir: nightly_tests stable: false frequency: nightly team: data cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances.yaml run: timeout: 7200 prepare: ' python setup_chaos.py --node-kill-interval 1200 --max-nodes-to-kill 3' script: RAY_DATA_PUSH_BASED_SHUFFLE=1 python dataset/sort.py --num-partitions=1000 --partition-size=1e9 wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances_gce.yaml - name: chaos_dataset_shuffle_sort_1tb group: data-tests working_dir: nightly_tests stable: false frequency: nightly team: data cluster: cluster_env: shuffle/shuffle_app_config_oom_disabled.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances.yaml run: timeout: 7200 prepare: 'python setup_chaos.py --node-kill-interval 900 --max-nodes-to-kill 3' script: python dataset/sort.py --num-partitions=1000 --partition-size=1e9 wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config_oom_disabled.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances_gce.yaml - name: chaos_dataset_shuffle_random_shuffle_1tb group: data-tests working_dir: nightly_tests stable: false frequency: nightly team: data cluster: # leave oom disabled as test is marked unstable at the moment. cluster_env: shuffle/shuffle_app_config_oom_disabled.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances.yaml run: timeout: 7200 prepare: ' python setup_chaos.py --node-kill-interval 600 --max-nodes-to-kill 2' script: python dataset/sort.py --num-partitions=1000 --partition-size=1e9 --shuffle wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config_oom_disabled.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances_gce.yaml - name: chaos_dataset_shuffle_push_based_random_shuffle_1tb group: data-tests working_dir: nightly_tests stable: false frequency: nightly team: data cluster: # leave oom disabled as test is marked unstable at the moment. cluster_env: shuffle/shuffle_app_config_oom_disabled.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances.yaml run: timeout: 7200 prepare: ' python setup_chaos.py --node-kill-interval 600 --max-nodes-to-kill 2' script: RAY_DATA_PUSH_BASED_SHUFFLE=1 python dataset/sort.py --num-partitions=1000 --partition-size=1e9 --shuffle wait_for_nodes: num_nodes: 20 variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: shuffle/shuffle_app_config_oom_disabled.yaml cluster_compute: shuffle/datasets_large_scale_compute_small_instances_gce.yaml ##################### # Observability tests ##################### - name: agent_stress_test group: core-observability-test working_dir: dashboard stable: true frequency: nightly team: core cluster: cluster_env: agent_stress_app_config.yaml cluster_compute: agent_stress_compute.yaml run: timeout: 14400 script: python mem_check.py --working-dir . variations: - __suffix__: aws - __suffix__: gce env: gce frequency: manual cluster: cluster_env: agent_stress_app_config.yaml cluster_compute: agent_stress_compute_gce.yaml - name: k8s_serve_ha_test group: k8s-test working_dir: k8s_tests stable: false frequency: nightly team: serve cluster: cluster_env: app_config.yaml cluster_compute: compute_tpl.yaml run: timeout: 28800 # 8h prepare: bash prepare.sh script: python run_gcs_ft_on_k8s.py - name: aws_cluster_launcher group: cluster-launcher-test working_dir: ../python/ray/autoscaler/ stable: true frequency: nightly team: core cluster: cluster_env: aws/tests/aws_config.yaml cluster_compute: aws/tests/aws_compute.yaml run: timeout: 1200 script: python launch_and_verify_cluster.py aws/tests/aws_cluster.yaml - name: aws_cluster_launcher_minimal group: cluster-launcher-test working_dir: ../python/ray/autoscaler/ stable: true frequency: nightly team: core cluster: cluster_env: aws/tests/aws_config.yaml cluster_compute: aws/tests/aws_compute.yaml run: timeout: 1200 script: python launch_and_verify_cluster.py aws/example-minimal.yaml - name: aws_cluster_launcher_full group: cluster-launcher-test working_dir: ../python/ray/autoscaler/ stable: true frequency: nightly team: core cluster: cluster_env: aws/tests/aws_config.yaml cluster_compute: aws/tests/aws_compute.yaml run: timeout: 1200 script: python launch_and_verify_cluster.py aws/example-full.yaml - name: gcp_cluster_launcher_minimal group: cluster-launcher-test working_dir: ../python/ray/autoscaler/ stable: true env: gce frequency: nightly team: core cluster: cluster_env: gcp/tests/gce_config.yaml cluster_compute: gcp/tests/single_node_32_cpu_gce.yaml run: timeout: 1200 script: python launch_and_verify_cluster.py gcp/example-minimal.yaml - name: gcp_cluster_launcher_full group: cluster-launcher-test working_dir: ../python/ray/autoscaler/ stable: true env: gce frequency: nightly team: core cluster: cluster_env: gcp/tests/gce_config.yaml cluster_compute: gcp/tests/single_node_32_cpu_gce.yaml run: timeout: 2400 script: python launch_and_verify_cluster.py gcp/example-full.yaml