o $i@sjddlmZmZddlmZmZddlmZddlm Z ddl m Z e \Z Z ZeGdddedZd S) )ABCMetaabstractmethod)AnyUnion) OldAPIStack) try_import_tf) TensorTypec@s|eZdZdZddZdeeefdefddZ deeefdefdd Z e deeefdefd d Z dedefd d Z dS)ScheduleaSchedule classes implement various time-dependent scheduling schemas. - Constant behavior. - Linear decay. - Piecewise decay. - Exponential decay. Useful for backend-agnostic rate/weight changes for learning rates, exploration epsilons, beta parameters for prioritized replay, loss weights decay, etc.. Each schedule can be called directly with the `t` (absolute time step) value and returns the value dependent on the Schedule and the passed time. cCs ||_dS)N) framework)selfr r _/home/ubuntu/veenaModal/venv/lib/python3.10/site-packages/ray/rllib/utils/schedules/schedule.py__init__s zSchedule.__init__treturncCs|jdvr ||S||S)zGenerates the value given a timestep (based on schedule's logic). Args: t: The time step. This could be a tf.Tensor. Returns: The calculated value depending on the schedule and `t`. )tf2tf)r _tf_value_op_valuer rr r r values  zSchedule.valuecC ||S)zCSimply calls self.value(t). Implemented to make Schedules callable.)rrr r r __call__,s zSchedule.__call__cCst)z Returns the value based on a time step input. Args: t: The time step. This could be a tf.Tensor. Returns: The calculated value depending on the schedule and `t`. )NotImplementedErrorrr r r r0s zSchedule._valuecCr)z Returns the tf-op that calculates the value based on a time step input. Args: t: The time step op (int tf.Tensor). Returns: The calculated value depending on the schedule and `t`. )rrr r r r=s zSchedule._tf_value_opN)__name__ __module__ __qualname____doc__rrintrrrrrrrr r r r r s  r ) metaclassN)abcrrtypingrrray.rllib.utils.annotationsrray.rllib.utils.frameworkrray.rllib.utils.typingrtf1rtfvr r r r r s