Struct CheckpointCallback 

pub struct CheckpointCallback {
    pub checkpoint_dir: PathBuf,
    pub save_frequency: usize,
    pub save_best_only: bool,
    pub keep_top_k: Option<usize>,
    /* private fields */
}

Callback for model checkpointing with auto-cleanup.

Fields

checkpoint_dir: PathBuf

Directory to save checkpoints.

save_frequency: usize

Frequency of checkpointing (every N epochs).

save_best_only: bool

Whether to save only the best model.

keep_top_k: Option<usize>

Maximum number of checkpoints to keep (None = keep all).

Implementations

impl CheckpointCallback

pub fn new( checkpoint_dir: PathBuf, save_frequency: usize, save_best_only: bool, ) -> Self

Create a new checkpoint callback.

pub fn with_cleanup( checkpoint_dir: PathBuf, save_frequency: usize, save_best_only: bool, keep_top_k: usize, ) -> Self

Create a new checkpoint callback with auto-cleanup.

This will automatically delete old checkpoints when the number exceeds keep_top_k, keeping only the checkpoints with the best (lowest) validation loss.

Arguments

• checkpoint_dir - Directory to save checkpoints
• save_frequency - Save every N epochs
• save_best_only - Only save when validation loss improves
• keep_top_k - Maximum number of checkpoints to keep (keeps best by validation loss)

Example

use tensorlogic_train::CheckpointCallback;
use std::path::PathBuf;

// Keep only the top 5 best checkpoints
let callback = CheckpointCallback::with_cleanup(
    PathBuf::from("/tmp/checkpoints"),
    1,    // save every epoch
    false, // save all, not just best
    5     // keep top 5
);

pub fn num_saved_checkpoints(&self) -> usize

Get the number of saved checkpoints being tracked.

pub fn cleanup_checkpoints(&mut self) -> TrainResult<usize>

Manually cleanup checkpoints, keeping only the top-k best.

This can be called manually to trigger cleanup if you’ve changed the keep_top_k setting.

Trait Implementations

impl Callback for CheckpointCallback

fn on_epoch_end( &mut self, epoch: usize, state: &TrainingState, ) -> TrainResult<()>

Called at the end of an epoch.

fn on_train_begin(&mut self, _state: &TrainingState) -> TrainResult<()>

Called at the beginning of training.

fn on_train_end(&mut self, _state: &TrainingState) -> TrainResult<()>

Called at the end of training.

fn on_epoch_begin( &mut self, _epoch: usize, _state: &TrainingState, ) -> TrainResult<()>

Called at the beginning of an epoch.

fn on_batch_begin( &mut self, _batch: usize, _state: &TrainingState, ) -> TrainResult<()>

Called at the beginning of a batch.

fn on_batch_end( &mut self, _batch: usize, _state: &TrainingState, ) -> TrainResult<()>

Called at the end of a batch.

fn on_validation_end(&mut self, _state: &TrainingState) -> TrainResult<()>

Called after validation.

fn should_stop(&self) -> bool

Check if training should stop early.