Trajectory - Tinker Documentation
tinker_cookbook.rl.Trajectory
class tinker_cookbook.rl.Trajectory()
A complete episode: a sequence of transitions from one agent in one environment.
A trajectory is produced by running an Env to completion. It contains all transitions (observation-action-reward triples) plus the final observation after the last action.
Fields:
- transitions ( list[ Transition])
- final_ob ( Observation)
- stop_reason ( str | None, default:
None)