Struct LearningCheckpointer 

pub struct LearningCheckpointer<M>
where
    M: LearnerModel,
{ /* private fields */ }

Used to create, delete, or load checkpoints of the training process.

Implementations

impl<M> LearningCheckpointer<M>

where M: LearnerModel,

pub fn new( model: AsyncCheckpointer<ModuleRecord>, optim: AsyncCheckpointer<OptimizerRecord>, lr_scheduler: AsyncCheckpointer<LrSchedulerRecord>, strategy: Box<dyn CheckpointingStrategy>, ) -> LearningCheckpointer<M>

Create a new learning checkpointer.

pub fn checkpoint( &mut self, learner: &Learner<M>, epoch: usize, store: &EventStoreClient, )

Create checkpoint for the training process.

pub fn load_checkpoint(&self, learner: Learner<M>, epoch: usize) -> Learner<M>

Load a training checkpoint.

No device is taken: checkpoints are device-free burnpack records (file-backed bytes). On load, the model keeps the device of the learner’s existing parameters, and the optimizer state is migrated to each parameter’s device on the next step. The training device is fixed earlier, when the learner’s model is created/forked.