derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
384 lines
17 KiB
JSON
384 lines
17 KiB
JSON
{
|
||
"$schema": "http://json-schema.org/draft-07/schema#",
|
||
"$id": "rccl-test-runner-config",
|
||
"title": "RCCL Test Runner Configuration",
|
||
"description": "Schema for RCCL test runner JSON configuration files",
|
||
"type": "object",
|
||
"additionalProperties": false,
|
||
"properties": {
|
||
|
||
"system_configurations": {
|
||
"description": "Human-readable metadata about this configuration",
|
||
"type": "object",
|
||
"additionalProperties": false,
|
||
"properties": {
|
||
"name": { "type": "string" },
|
||
"description": { "type": "string" }
|
||
}
|
||
},
|
||
|
||
"paths": {
|
||
"description": "Filesystem paths used by the runner. Values support ${VAR}, $VAR, and ${VAR:-default} expansion.",
|
||
"type": "object",
|
||
"additionalProperties": false,
|
||
"properties": {
|
||
"workdir": { "type": "string", "description": "RCCL source and build root (default: $PWD)" },
|
||
"rocm_path": { "type": "string", "description": "ROCm installation prefix (default: /opt/rocm)" },
|
||
"mpi_path": { "type": "string", "description": "MPI installation prefix" },
|
||
"test_binary_dir": { "type": "string", "description": "Directory containing test binaries; overrides the default build_dir/test/" }
|
||
}
|
||
},
|
||
|
||
"env_variables": {
|
||
"$ref": "#/$defs/env_variables_map",
|
||
"description": "Global environment variables applied to every test run"
|
||
},
|
||
|
||
"auto_detect_hosts": {
|
||
"type": "boolean",
|
||
"description": "When true, auto-detect MPI hosts from SLURM_NODELIST or similar"
|
||
},
|
||
|
||
"mpi_args": {
|
||
"description": "Base mpirun arguments that replace the built-in defaults. May be a string or list of strings applied to all tests, OR an object keyed by --system profile name (e.g. 'mellanox', 'ainic', 'thor2') whose values are strings or lists of strings.",
|
||
"oneOf": [
|
||
{ "$ref": "#/$defs/mpi_args_value" },
|
||
{
|
||
"type": "object",
|
||
"additionalProperties": { "$ref": "#/$defs/mpi_args_value" }
|
||
}
|
||
]
|
||
},
|
||
|
||
"system_env_variables": {
|
||
"description": "Per-system-profile environment variable overrides (keyed by profile name)",
|
||
"type": "object",
|
||
"additionalProperties": { "$ref": "#/$defs/env_variables_map" }
|
||
},
|
||
|
||
"build_configuration": {
|
||
"description": "Controls how librccl.so and test binaries are built via install.sh",
|
||
"type": "object",
|
||
"additionalProperties": false,
|
||
"properties": {
|
||
"install_flags": {
|
||
"type": "array",
|
||
"items": { "type": "string" },
|
||
"description": "install.sh command-line flags, e.g. ['--debug', '-t', '-l', '--no_clean']"
|
||
},
|
||
"cmake_options": {
|
||
"description": "Additional CMake options. Either a string ('-DFOO=BAR -DBAZ=1') or an object whose keys become -DKEY=VAL entries.",
|
||
"oneOf": [
|
||
{ "type": "string" },
|
||
{
|
||
"type": "object",
|
||
"additionalProperties": { "type": ["string", "number", "boolean"] }
|
||
}
|
||
]
|
||
},
|
||
"env_variables": {
|
||
"$ref": "#/$defs/env_variables_map",
|
||
"description": "Environment variables set only during the build step"
|
||
},
|
||
"parallel_jobs": {
|
||
"type": "integer",
|
||
"minimum": 1,
|
||
"description": "Number of parallel compile jobs passed to install.sh via -j"
|
||
},
|
||
"generator": {
|
||
"type": "string",
|
||
"description": "CMake generator, e.g. 'Unix Makefiles' or 'Ninja'"
|
||
}
|
||
}
|
||
},
|
||
|
||
"rccl_tests_build_configuration": {
|
||
"description": "Controls how the rccl-tests perf binaries (all_reduce_perf, etc.) are built. Absent section => step skipped.",
|
||
"type": "object",
|
||
"additionalProperties": false,
|
||
"properties": {
|
||
"enabled": {
|
||
"type": "boolean",
|
||
"description": "Set to false to skip building rccl-tests. Default: true when the section is present."
|
||
},
|
||
"source_dir": {
|
||
"type": "string",
|
||
"description": "Path to the rccl-tests checkout. Supports ~ and $VAR expansion. Default: $workdir/rccl-tests."
|
||
},
|
||
"install_script": {
|
||
"type": "string",
|
||
"description": "Name of the build script inside source_dir. Default: 'install.sh'."
|
||
},
|
||
"build_command": {
|
||
"description": "Explicit build command (string run via shell, or argv array). Overrides install_script + install_flags.",
|
||
"oneOf": [
|
||
{ "type": "string" },
|
||
{ "type": "array", "items": { "type": "string" } }
|
||
]
|
||
},
|
||
"install_flags": {
|
||
"type": "array",
|
||
"items": { "type": "string" },
|
||
"description": "Flags passed to install_script, e.g. ['--mpi']. Supports ~ and $VAR expansion."
|
||
},
|
||
"rccl_home": {
|
||
"type": "string",
|
||
"description": "RCCL install/build prefix to link rccl-tests against. Default: the RCCL build_dir."
|
||
},
|
||
"env_variables": {
|
||
"$ref": "#/$defs/env_variables_map",
|
||
"description": "Environment variables set only during the rccl-tests build step"
|
||
}
|
||
}
|
||
},
|
||
|
||
"test_configurations": {
|
||
"description": "Named test configuration blocks. Each key is a configuration name referenced by test_suites or via 'extends'.",
|
||
"type": "object",
|
||
"additionalProperties": { "$ref": "#/$defs/test_configuration" }
|
||
},
|
||
|
||
"test_suites": {
|
||
"description": "Ordered list of test suites to execute",
|
||
"type": "array",
|
||
"items": { "$ref": "#/$defs/test_suite" }
|
||
}
|
||
},
|
||
|
||
"$defs": {
|
||
|
||
"env_variables_map": {
|
||
"description": "Key/value pairs of environment variables. Values are always strings (use '' to unset).",
|
||
"type": "object",
|
||
"additionalProperties": { "type": "string" }
|
||
},
|
||
|
||
"mpi_args_value": {
|
||
"description": "Extra mpirun arguments expressed as either a single string ('--bind-to none --oversubscribe') or a list of strings (['--bind-to none', '--oversubscribe']).",
|
||
"oneOf": [
|
||
{ "type": "string" },
|
||
{ "type": "array", "items": { "type": "string" } }
|
||
]
|
||
},
|
||
|
||
"test_defaults": {
|
||
"description": "Fields that can appear at the test_configuration level as defaults and be overridden per-test",
|
||
"type": "object",
|
||
"properties": {
|
||
"is_gtest": {
|
||
"type": "boolean",
|
||
"description": "True → GTest binary using --gtest_filter; false → plain executable with direct args. Default: true."
|
||
},
|
||
"binary": {
|
||
"type": "string",
|
||
"description": "Test binary name or path. Relative names are resolved under build_dir/test/ (or test_binary_dir if set). Supports ~ and $VAR expansion."
|
||
},
|
||
"num_ranks": {
|
||
"description": "Total MPI ranks. 1 → no mpirun wrapper. The string \"auto\" resolves to (detected GPUs per node × num_nodes), i.e. use all GPUs.",
|
||
"oneOf": [
|
||
{ "type": "integer", "minimum": 1 },
|
||
{ "type": "string", "enum": ["auto"] }
|
||
]
|
||
},
|
||
"num_nodes": {
|
||
"type": "integer",
|
||
"minimum": 1,
|
||
"description": "Number of nodes across which ranks are distributed"
|
||
},
|
||
"num_gpus": {
|
||
"description": "GPUs per node; controls --map-by ppr:{num_gpus}:node for multi-node runs. The string \"auto\" (also the default when omitted) resolves to the detected GPUs per node.",
|
||
"oneOf": [
|
||
{ "type": "integer", "minimum": 1 },
|
||
{ "type": "string", "enum": ["auto"] }
|
||
]
|
||
},
|
||
"timeout": {
|
||
"type": "integer",
|
||
"minimum": 0,
|
||
"description": "Timeout in seconds. 0 = unlimited."
|
||
},
|
||
"is_pytest": {
|
||
"type": "boolean",
|
||
"description": "True → run a pytest harness (in test_dir) instead of a gtest/plain binary; results are derived from a JUnit report. Default: false."
|
||
},
|
||
"test_dir": {
|
||
"type": "string",
|
||
"description": "For pytest tests: directory containing the harness (absolute, or relative to workdir). Supports ~ and $VAR expansion."
|
||
},
|
||
"python_bin": {
|
||
"type": "string",
|
||
"description": "For pytest tests: interpreter to use (default: <test_dir>/venv else python3, or the venv created by setup_venv)."
|
||
},
|
||
"setup_venv": {
|
||
"type": "boolean",
|
||
"description": "For pytest tests: create/reuse a venv and install requirements before running. Default: false."
|
||
},
|
||
"venv_dir": {
|
||
"type": "string",
|
||
"description": "For pytest tests with setup_venv: venv location (default: <test_dir>/venv)."
|
||
},
|
||
"requirements": {
|
||
"type": "string",
|
||
"description": "For pytest tests with setup_venv: requirements file to install (default: <test_dir>/requirements.txt)."
|
||
}
|
||
}
|
||
},
|
||
|
||
"test_configuration": {
|
||
"description": "A named test configuration. May extend another configuration and define a list of tests.",
|
||
"type": "object",
|
||
"additionalProperties": false,
|
||
"properties": {
|
||
"extends": {
|
||
"description": "Name (or list of names) of parent configuration(s) to inherit from",
|
||
"oneOf": [
|
||
{ "type": "string" },
|
||
{ "type": "array", "items": { "type": "string" }, "minItems": 1 }
|
||
]
|
||
},
|
||
"is_gtest": { "$ref": "#/$defs/test_defaults/properties/is_gtest" },
|
||
"binary": { "$ref": "#/$defs/test_defaults/properties/binary" },
|
||
"num_ranks": { "$ref": "#/$defs/test_defaults/properties/num_ranks" },
|
||
"num_nodes": { "$ref": "#/$defs/test_defaults/properties/num_nodes" },
|
||
"num_gpus": { "$ref": "#/$defs/test_defaults/properties/num_gpus" },
|
||
"timeout": { "$ref": "#/$defs/test_defaults/properties/timeout" },
|
||
"is_pytest": { "$ref": "#/$defs/test_defaults/properties/is_pytest" },
|
||
"test_dir": { "$ref": "#/$defs/test_defaults/properties/test_dir" },
|
||
"python_bin": { "$ref": "#/$defs/test_defaults/properties/python_bin" },
|
||
"setup_venv": { "$ref": "#/$defs/test_defaults/properties/setup_venv" },
|
||
"venv_dir": { "$ref": "#/$defs/test_defaults/properties/venv_dir" },
|
||
"requirements": { "$ref": "#/$defs/test_defaults/properties/requirements" },
|
||
"env_variables": {
|
||
"$ref": "#/$defs/env_variables_map",
|
||
"description": "Environment variables applied to every test in this configuration (merged on top of global env_variables)"
|
||
},
|
||
"rerun_env_variables": {
|
||
"$ref": "#/$defs/env_variables_map",
|
||
"description": "Additional environment variables merged in when a failed test is rerun (requires --rerun-failed)"
|
||
},
|
||
"args": {
|
||
"type": "array",
|
||
"items": { "type": "string" },
|
||
"description": "Extra command-line arguments appended to the test binary invocation. Lists are de-duplicated and appended when configs are merged via 'extends'."
|
||
},
|
||
"mpi_args": {
|
||
"$ref": "#/$defs/mpi_args_value",
|
||
"description": "Extra mpirun arguments for tests in this configuration, appended after the base mpi_args. Accepts a string or list of strings; lists are de-duplicated and appended when configs are merged via 'extends'."
|
||
},
|
||
"command_args": {
|
||
"type": "string",
|
||
"description": "Shared base command-line arguments for this configuration. A test's own 'command_args' is appended to this base at execution time (see test_executor.run_test). Inherited via 'extends': a child that omits command_args takes the parent's value."
|
||
},
|
||
"description": {
|
||
"type": "string",
|
||
"description": "Human-readable description of this configuration block (not interpreted by the runner)"
|
||
},
|
||
"tests": {
|
||
"type": "array",
|
||
"items": { "$ref": "#/$defs/test_item" },
|
||
"description": "Individual test cases belonging to this configuration"
|
||
}
|
||
}
|
||
},
|
||
|
||
"test_item": {
|
||
"description": "A single test case within a test_configuration",
|
||
"type": "object",
|
||
"required": ["name"],
|
||
"additionalProperties": false,
|
||
"properties": {
|
||
"name": {
|
||
"type": "string",
|
||
"description": "Unique test identifier used for --test-name filtering (supports glob patterns)"
|
||
},
|
||
"description": {
|
||
"type": "string",
|
||
"description": "Human-readable description shown in output and logs"
|
||
},
|
||
"is_gtest": { "$ref": "#/$defs/test_defaults/properties/is_gtest" },
|
||
"binary": { "$ref": "#/$defs/test_defaults/properties/binary" },
|
||
"test_binary_dir": {
|
||
"type": "string",
|
||
"description": "Per-test override of the directory in which binary is looked up"
|
||
},
|
||
"test_filter": {
|
||
"type": "string",
|
||
"description": "For GTest binaries: passed as --gtest_filter=<value>. For non-GTest: passed as a plain argument."
|
||
},
|
||
"command_args": {
|
||
"type": "string",
|
||
"description": "Additional command-line arguments appended after the binary (and after test_filter for non-GTest)"
|
||
},
|
||
"num_ranks": { "$ref": "#/$defs/test_defaults/properties/num_ranks" },
|
||
"num_nodes": { "$ref": "#/$defs/test_defaults/properties/num_nodes" },
|
||
"num_gpus": { "$ref": "#/$defs/test_defaults/properties/num_gpus" },
|
||
"timeout": { "$ref": "#/$defs/test_defaults/properties/timeout" },
|
||
"is_pytest": { "$ref": "#/$defs/test_defaults/properties/is_pytest" },
|
||
"test_dir": { "$ref": "#/$defs/test_defaults/properties/test_dir" },
|
||
"python_bin": { "$ref": "#/$defs/test_defaults/properties/python_bin" },
|
||
"setup_venv": { "$ref": "#/$defs/test_defaults/properties/setup_venv" },
|
||
"venv_dir": { "$ref": "#/$defs/test_defaults/properties/venv_dir" },
|
||
"requirements": { "$ref": "#/$defs/test_defaults/properties/requirements" },
|
||
"env_variables": {
|
||
"$ref": "#/$defs/env_variables_map",
|
||
"description": "Test-specific environment variables (highest priority, merged on top of all other levels)"
|
||
},
|
||
"rerun_env_variables": {
|
||
"$ref": "#/$defs/env_variables_map",
|
||
"description": "Per-test additional environment variables used only on rerun"
|
||
},
|
||
"mpi_args": {
|
||
"$ref": "#/$defs/mpi_args_value",
|
||
"description": "Per-test extra mpirun arguments (MPI tests only), appended after the base and suite/configuration mpi_args. Accepts a string or list of strings."
|
||
}
|
||
}
|
||
},
|
||
|
||
"test_suite": {
|
||
"description": "A suite entry that binds a test_configuration to a named run with optional overrides",
|
||
"type": "object",
|
||
"required": ["name", "config"],
|
||
"additionalProperties": false,
|
||
"properties": {
|
||
"name": {
|
||
"type": "string",
|
||
"description": "Human-readable suite name used for --suite-name filtering"
|
||
},
|
||
"description": {
|
||
"type": "string",
|
||
"description": "Optional longer description of what this suite exercises"
|
||
},
|
||
"config": {
|
||
"type": "string",
|
||
"description": "Name of the test_configuration to use for this suite"
|
||
},
|
||
"enabled": {
|
||
"type": "boolean",
|
||
"description": "When false the suite is skipped unless targeted by --suite-name. Default: true."
|
||
},
|
||
"is_gtest": { "$ref": "#/$defs/test_defaults/properties/is_gtest" },
|
||
"binary": { "$ref": "#/$defs/test_defaults/properties/binary" },
|
||
"num_ranks": { "$ref": "#/$defs/test_defaults/properties/num_ranks" },
|
||
"num_nodes": { "$ref": "#/$defs/test_defaults/properties/num_nodes" },
|
||
"num_gpus": { "$ref": "#/$defs/test_defaults/properties/num_gpus" },
|
||
"timeout": { "$ref": "#/$defs/test_defaults/properties/timeout" },
|
||
"is_pytest": { "$ref": "#/$defs/test_defaults/properties/is_pytest" },
|
||
"test_dir": { "$ref": "#/$defs/test_defaults/properties/test_dir" },
|
||
"python_bin": { "$ref": "#/$defs/test_defaults/properties/python_bin" },
|
||
"setup_venv": { "$ref": "#/$defs/test_defaults/properties/setup_venv" },
|
||
"venv_dir": { "$ref": "#/$defs/test_defaults/properties/venv_dir" },
|
||
"requirements": { "$ref": "#/$defs/test_defaults/properties/requirements" },
|
||
"env_variables": {
|
||
"$ref": "#/$defs/env_variables_map",
|
||
"description": "Suite-level environment variable overrides (between configuration-level and test-level in priority)"
|
||
},
|
||
"mpi_args": {
|
||
"$ref": "#/$defs/mpi_args_value",
|
||
"description": "Suite-level extra mpirun arguments, appended after the base and configuration mpi_args (and before per-test mpi_args). Accepts a string or list of strings."
|
||
}
|
||
}
|
||
}
|
||
|
||
}
|
||
}
|