trinity.algorithm.sample_strategy package

trinity.algorithm.sample_strategy package#

class trinity.algorithm.sample_strategy.SampleStrategy(buffer_config: BufferConfig, **kwargs)[source]#

Bases: ABC

set_model_version_metric(exp_list: List[Experience], metrics: Dict)[source]#

abstract async sample(step: int) → Tuple[Experiences, Dict, List][source]#

Sample data from buffer.

Parameters:: step (int) – The step number of current step.
Returns:: The sampled Experiences data. Dict: Metrics for logging. List: Representative data for logging.
Return type:: Experiences

abstract classmethod default_args() → dict[source]#: Get the default arguments of the sample strategy.

abstract state_dict() → dict[source]#: Get the state dict of the sample strategy.

abstract load_state_dict(state_dict: dict) → None[source]#: Load the state dict of the sample strategy.

class trinity.algorithm.sample_strategy.DefaultSampleStrategy(buffer_config: BufferConfig, **kwargs)[source]#

async sample(step: int, **kwargs) → Tuple[Experiences, Dict, List][source]#

Sample data from buffer.

Parameters:: step (int) – The step number of current step.
Returns:: The sampled Experiences data. Dict: Metrics for logging. List: Representative data for logging.
Return type:: Experiences

classmethod default_args() → dict[source]#: Get the default arguments of the sample strategy.

load_state_dict(state_dict: dict) → None[source]#: Load the state dict of the sample strategy.

class trinity.algorithm.sample_strategy.WarmupSampleStrategy(buffer_config: BufferConfig, **kwargs)[source]#

The warmup sample strategy. Deprecated, keep this class for backward compatibility only. Please use DefaultSampleStrategy instead.

class trinity.algorithm.sample_strategy.MixSampleStrategy(buffer_config: BufferConfig, **kwargs)[source]#

The default sample strategy.

async sample(step: int) → Tuple[Experiences, Dict, List][source]#

Sample data from buffer.

Parameters:: step (int) – The step number of current step.
Returns:: The sampled Experiences data. Dict: Metrics for logging. List: Representative data for logging.
Return type:: Experiences

classmethod default_args() → Dict[source]#: Get the default arguments of the sample strategy.

load_state_dict(state_dict: dict) → None[source]#: Load the state dict of the sample strategy.