trinity.common.workflows.agentscope_workflow module#

class trinity.common.workflows.agentscope_workflow.AgentScopeWorkflowAdapter(*, task: Task, model: ModelWrapper, auxiliary_models: List[ModelWrapper] | None = None)[source]#

Bases: Workflow

Adapter to wrap a agentscope trainable workflow function into a Trinity Workflow. Only for agentscope versions between 1.0.7 and 1.0.11. For agentscope >= 1.0.12, please use AgentScopeWorkflowAdapterV1.

is_async: bool = True#
__init__(*, task: Task, model: ModelWrapper, auxiliary_models: List[ModelWrapper] | None = None)[source]#

Initialize the adapter with the task and model.

construct_experiences(reward: float) List[Experience][source]#

Construct experiences from the agent’s interaction history.

Parameters:

reward (float) – The reward value to assign to each experience.

Returns:

A list of Experience objects.

Return type:

List

async run_async() List[Experience][source]#

Run the workflow asynchronously and return experiences.

class trinity.common.workflows.agentscope_workflow.AgentScopeWorkflowAdapterV1(*, task: Task, model: ModelWrapper, auxiliary_models: List[ModelWrapper] | None = None)[source]#

Bases: Workflow

A more general adapter to wrap agentscope trainable workflow and judge functions into a Trinity Workflow. Only for agentscope versions >= 1.0.12.

is_async: bool = True#
__init__(*, task: Task, model: ModelWrapper, auxiliary_models: List[ModelWrapper] | None = None)[source]#

Initialize the adapter with the task and model.

construct_experiences(reward: float, metrics: Dict) List[Experience][source]#

Construct experiences from the agent’s interaction history.

Parameters:
  • reward (float) – The reward value to assign to each experience.

  • metrics (Dict) – A dictionary of metrics to be attached to the last experience.

Returns:

A list of Experience objects.

Return type:

List

async run_async() List[Experience][source]#

Run the workflow asynchronously and return experiences.