nip.artifact_logger.ArtifactLogger

class nip.artifact_logger.ArtifactLogger(settings: ExperimentSettings, agents: dict[str, Agent])

Samples rollouts from an environment and saves them to W&B.

Parameters:

• settings (ExperimentSettings) – The settings of the experiment.

• agents (dict[str, Agent]) – The agents in the experiment.

Methods Summary

__init__(settings, agents)
_sample_and_save_rollouts(data, iteration)	Sample rollouts from the given data and save them to W&B.
_save_checkpoint()	Save the agent models to W&B.
log(data, iteration)	Log artifacts to W&B if it's time to do so.

Methods

__init__(settings: ExperimentSettings, agents: dict[str, Agent])

_sample_and_save_rollouts(data: TensorDict, iteration: int)

Sample rollouts from the given data and save them to W&B.

Parameters:

• data (TensorDict) – The data to sample rollouts from.

• iteration (int) – The iteration of training we are on.

_save_checkpoint()

Save the agent models to W&B.

log(data: TensorDict, iteration: int)

Log artifacts to W&B if it’s time to do so.

Parameters:

• data (TensorDict) – The data sampled most recently from the data collector.

• iteration (int) – The iteration of training we are on.