evorl.recorders

Package Contents

Classes

ChainRecorder	Container for multiple recorders.
LogRecorder	Log file recorder.
Recorder	A Recorder Interface.
WandbRecorder	Recorder for Weights & Biases.

Functions

add_prefix	Add prefix to the keys of a dictionary.
get_1d_array	Get statistics of a 1D array.
get_1d_array_statistics	Get raw value and statistics of a 1D array.

API

class evorl.recorders.ChainRecorder(recorders: collections.abc.Sequence[evorl.recorders.recorder.Recorder])

Bases: evorl.recorders.recorder.Recorder

Container for multiple recorders.

add_recorder(recorder: evorl.recorders.recorder.Recorder) → None

close() → None

init() → None

write(data: collections.abc.Mapping[str, Any], step: int | None = None) → None

class evorl.recorders.LogRecorder(log_path: str, console: bool = True)

Bases: evorl.recorders.recorder.Recorder

Log file recorder.

close() → None

init() → None

write(data: collections.abc.Mapping[str, Any], step: int | None = None) → None

class evorl.recorders.Recorder

Bases: abc.ABC

A Recorder Interface.

abstract close() → None

Finalize the recorder.

abstract init() → None

Initialize the recorder.

abstract write(data: collections.abc.Mapping[str, Any], step: int | None = None) → None

Write data to the recorder.

class evorl.recorders.WandbRecorder(*, project, name, config, tags, path, **wandb_kwargs)

Bases: evorl.recorders.recorder.Recorder

Recorder for Weights & Biases.

close()

init() → None

write(data: collections.abc.Mapping[str, Any], step: int | None = None) → None

evorl.recorders.add_prefix(data: dict, prefix: str)

Add prefix to the keys of a dictionary.

evorl.recorders.get_1d_array(data)

Get statistics of a 1D array.

Similar to get_1d_array_statistics, but instead of recording histogram, WandB will record the raw data.

evorl.recorders.get_1d_array_statistics(data, histogram=False)

Get raw value and statistics of a 1D array.

Helper function for logging in WandB.

Parameters:

• data – 1D numpy array. If data has multiple dimensions, it will be viewed as flattened.

• histogram – If True, return raw data in pd.Series, which will be futher converted to histogram in WandBRecorder.

Returns:

A dictionary containing min, max, mean, and optional raw data.