composer.core#

Modules

composer.core.algorithm

Base class for algorithms that improve model's quality or efficiency.

composer.core.callback

Base module for callbacks.

composer.core.data_spec

Specifications for operating and training on data.

composer.core.engine

Engine is a coordinator for running algorithms and resolving ordering conflicts among them for composition.

composer.core.evaluator

A wrapper for a dataloader to include metrics that apply to a specific dataset.

composer.core.event

Events represent specific points in the training loop where an Algorithm and Callback can run.

composer.core.logging

Base classes, functions, and variables for logging.

composer.core.precision

Enum class for the numerical precision to be used by the model.

composer.core.serializable

Serialization interface used by checkpointing.

composer.core.state

The state of the trainer.

composer.core.time

Utilities to track training progress in terms of epochs, batches, samples, and tokens.

composer.core.types

Reference for common types used throughout the composer library.

Central components used by other modules.

Central parts of composer such as Engine, base class for critical components such as Algorithm and Callback and other useful functionality such as Logger and Timer are implemented under core.

Classes

Algorithm

Base class for algorithms.

Callback

Base class for callbacks.

DataSpec

Specifications for operating and training on data.

Engine

Coordinator for running algorithms and resolving ordering conflicts among them for composition.

Evaluator

A wrapper for a dataloader to include metrics that apply to a specific dataset.

Event

Enum to represent events in the training loop.

Logger

Logger routes metrics to the LoggerCallback.

State

The state of the trainer.

Time

Time represents static durations of training time or points in the training process in terms of a TimeUnit enum (epochs, batches, samples, tokens, or duration).

TimeUnit

Enum class to represent units of time for the training process.

Timer

Timer tracks the current training progress, in terms of epochs, batches, samples, and tokens.

Timestamp

Timestamp represents a snapshot of Timer.

Trace

Record of an algorithm's execution.