nemo_gym.benchmarks

Benchmark discovery and preparation utilities.

Module Contents

Classes

Name	Description
`BenchmarkConfig`	-

Functions

Name	Description
`_benchmark_config_name`	The name of the benchmark config, given its path relative to `benchmarks/`, sans `.yaml`.
`_benchmark_config_paths`	Sorted config paths under one dir that declare a benchmark, discovered by content.
`_discover_benchmarks_in_dir`	Map benchmark name -> :class:`BenchmarkConfig` for every benchmark config under one dir.
`_is_benchmark_config`	True if the config declares a `type: benchmark` dataset anywhere in its structure.
`discover_benchmarks`	Map benchmark name -> :class:`BenchmarkConfig` for every discoverable benchmark config.

Data

BENCHMARKS_DIR

BENCHMARKS_SUBDIR

__getattr__

API

class nemo_gym.benchmarks.BenchmarkConfig()

Bases: BaseModel

agent_name

str

dataset

BenchmarkDatasetConfig

name

str

num_repeats

int

path

Path

nemo_gym.benchmarks.BenchmarkConfig.from_config_path(
    config_path: pathlib.Path,
    strict: bool = True
) -> typing.Optional[nemo_gym.benchmarks.BenchmarkConfig]

classmethod

nemo_gym.benchmarks.BenchmarkConfig.from_initial_config_dict(
    path: pathlib.Path,
    initial_config_dict: omegaconf.DictConfig,
    strict: bool = True
) -> typing.Optional[nemo_gym.benchmarks.BenchmarkConfig]

classmethod

nemo_gym.benchmarks._benchmark_config_name(
    rel_config_path: pathlib.Path
) -> str

The name of the benchmark config, given its path relative to benchmarks/, sans .yaml.

This is the identity we key benchmarks by, so a listed benchmark is always a valid --benchmark argument.

nemo_gym.benchmarks._benchmark_config_paths(
    benchmarks_dir: pathlib.Path
) -> typing.List[pathlib.Path]

Sorted config paths under one dir that declare a benchmark, discovered by content.

A config is a benchmark iff it declares a type: benchmark dataset, regardless of filename, so we scan every yaml. :func:_is_benchmark_config is a cheap prefilter (pay the resolve cost only on real candidates) that also catches non-config.yaml names like tau2’s configs/*.yaml. Empty if dir missing.

nemo_gym.benchmarks._discover_benchmarks_in_dir(
    benchmarks_dir: pathlib.Path
) -> typing.Dict[str, nemo_gym.benchmarks.BenchmarkConfig]

Map benchmark name -> :class:BenchmarkConfig for every benchmark config under one dir.

nemo_gym.benchmarks._is_benchmark_config(
    config_path: pathlib.Path
) -> bool

True if the config declares a type: benchmark dataset anywhere in its structure.

A raw single-file parse (no config_paths/interpolation resolution), so it’s format-agnostic and can’t fail on includes. An unparseable file is kept (returns True) so the resolve step surfaces a diagnostic.

nemo_gym.benchmarks.discover_benchmarks() -> typing.Dict[str, nemo_gym.benchmarks.BenchmarkConfig]

Map benchmark name -> :class:BenchmarkConfig for every discoverable benchmark config.

Scans the benchmarks/ subdir of every :func:~nemo_gym.discovery.component_search_roots root (NEMO_GYM_EXTRA_ROOTS + cwd + built-ins), merged so user benchmarks shadow same-named built-ins.

nemo_gym.benchmarks.BENCHMARKS_DIR = PARENT_DIR / BENCHMARKS_SUBDIR

nemo_gym.benchmarks.BENCHMARKS_SUBDIR = 'benchmarks'

nemo_gym.benchmarks.__getattr__ = moved_attr_getter(__name__, {'list_benchmarks': 'nemo_gym.cli.eval', 'PrepareBen...