aitune.utils.monitoring.hardware_metrics.metrics

View as Markdown

Metrics providers for hardware metrics.

Module Contents

Classes

NameDescription
AbstractHardwareMetricProviderAbstract base class for hardware metric providers.
HostMetricProviderHost metric provider.
NvmlMetricProviderNvml metric provider.
TorchMetricProviderTorch metric provider.

API

class aitune.utils.monitoring.hardware_metrics.metrics.AbstractHardwareMetricProvider()
Abstract

Abstract base class for hardware metric providers.

aitune.utils.monitoring.hardware_metrics.metrics.AbstractHardwareMetricProvider.get_metrics() -> dict[str, typing.Any]
abstract

Gets the metrics.

class aitune.utils.monitoring.hardware_metrics.metrics.HostMetricProvider()

Bases: AbstractHardwareMetricProvider

Host metric provider.

aitune.utils.monitoring.hardware_metrics.metrics.HostMetricProvider.get_metrics() -> dict[str, typing.Any]

Gets the metrics.

class aitune.utils.monitoring.hardware_metrics.metrics.NvmlMetricProvider()

Bases: AbstractHardwareMetricProvider

Nvml metric provider.

handles
aitune.utils.monitoring.hardware_metrics.metrics.NvmlMetricProvider.close()

Closes the NvmlMetricProvider.

aitune.utils.monitoring.hardware_metrics.metrics.NvmlMetricProvider.get_metrics() -> dict[str, typing.Any]

Gets the metrics.

class aitune.utils.monitoring.hardware_metrics.metrics.TorchMetricProvider()

Bases: AbstractHardwareMetricProvider

Torch metric provider.

aitune.utils.monitoring.hardware_metrics.metrics.TorchMetricProvider.get_metrics() -> dict[str, typing.Any]

Gets the metrics.