tracel-ai/burn · error

Can delete optimizer checkpoint.

Error message

Can delete optimizer checkpoint.

What it means

Same checkpointing flow as the model delete: the optimizer checkpointer's `delete(epoch)` is unwrapped with expect() and panics with this message when the optimizer checkpoint for that epoch cannot be deleted (file missing or IO error).

Source

Thrown at crates/burn-train/src/learner/base.rs:170

            lr_scheduler,
            strategy,
            _phantom: PhantomData,
        }
    }

    /// Create checkpoint for the training process.
    pub fn checkpoint(&mut self, learner: &Learner<M>, epoch: usize, store: &EventStoreClient) {
        let actions = self.strategy.checkpointing(epoch, store);

        for action in actions {
            match action {
                CheckpointingAction::Delete(epoch) => {
                    self.model
                        .delete(epoch)
                        .expect("Can delete model checkpoint.");
                    self.optim
                        .delete(epoch)
                        .expect("Can delete optimizer checkpoint.");
                    self.lr_scheduler
                        .delete(epoch)
                        .expect("Can delete learning rate scheduler checkpoint.");
                }
                CheckpointingAction::Save => {
                    self.model
                        .save(epoch, learner.model.clone().into_record())
                        .expect("Can save model checkpoint.");
                    self.optim
                        .save(epoch, learner.optim.to_record())
                        .expect("Can save optimizer checkpoint.");
                    self.lr_scheduler
                        .save(epoch, learner.lr_scheduler.to_record())
                        .expect("Can save learning rate scheduler checkpoint.");
                }
            }
        }
    }

View on GitHub (pinned to d16f7ba2ed)

Solutions

1. Verify the optimizer checkpoint file for the epoch exists in the checkpoint directory and is writable.
2. Keep model/optimizer/scheduler checkpoints as an atomic set — restore or delete all three together.
3. Avoid external cleanup of individual checkpoint files while training is running.
4. Use a dedicated checkpoint directory per run to prevent concurrent deletion.
5. If your workflow deletes files externally, disable the Delete actions or make your wrapper tolerate missing epochs.

Example fix

// before: deleting epoch 12 after manual cleanup removed only optim-12
// after: delete the whole epoch as a unit or check first
for prefix in ["model", "optim", "scheduler"] {
    assert!(Path::new(dir).join(format!("{prefix}-{epoch}")).exists(), "{prefix} checkpoint missing");
}
checkpointer.checkpoint(actions);

let optim_path = Path::new(&checkpoint_dir).join(format!("optim-{epoch}"));
assert!(optim_path.exists(), "optimizer checkpoint for epoch {epoch} not found before delete");

fn has_full_checkpoint(dir: &str, epoch: usize) -> bool {
    ["model", "optim", "scheduler"].iter().all(|p| Path::new(dir).join(format!("{p}-{epoch}")).exists())
}

let result = std::panic::catch_unwind(AssertUnwindSafe(|| checkpointer.checkpoint(actions)));
if result.is_err() { /* restore or skip */ }

When it happens

Trigger: Applying CheckpointingAction::Delete(epoch) where the optimizer checkpoint file for `epoch` is absent, already removed, or the recorder lacks permission/disk access to remove it.

Common situations: Partial checkpoint sets (model file present but optimizer file deleted by disk cleanup or an interrupted save); wrong checkpoint directory; concurrent jobs deleting each other's files.

Related errors

• Can save optimizer checkpoint.
• Failed to load module from file
• Can delete model checkpoint.
• Can delete learning rate scheduler checkpoint.
• Can save model checkpoint.

AI-assisted analysis of tracel-ai/burn@d16f7ba2ed (2026-09-05). Data as JSON: /api/errors/03419bb2bffe0801. Report an issue: GitHub.