tlm_adjoint.checkpoint_schedules.mixed

Module Contents

class tlm_adjoint.checkpoint_schedules.mixed.MixedCheckpointSchedule(max_n, snapshots, *, storage='disk')

A checkpointing schedule which mixes storage of forward restart data and non-linear dependency data in checkpointing units. Assumes that the data required to restart the forward has the same size as the data required to advance the adjoint over a step.

Described in

• James R. Maddison, ‘On the implementation of checkpointing with high-level algorithmic differentiation’, https://arxiv.org/abs/2305.09568v1, 2023

Offline, one adjoint calculation permitted.

Parameters:

• max_n – The number of forward steps in the initial forward calculation.

• snapshots – The number of available checkpointing units.

• storage – Checkpointing unit storage location. Either ‘RAM’ or ‘disk’.

property is_exhausted

Whether the schedule has concluded. Note that some schedules permit multiple adjoint calculation, and may never conclude.

property uses_disk_storage

Whether the schedule may use disk storage. If False then no disk storage is required.

iter()

A generator which should be overridden in derived classes in order to define a checkpointing schedule.