tracel-ai/burn · error

The sequential scheduler step counter overflowed

Error message

The sequential scheduler step counter overflowed

What it means

Overflow guard in `SequentialLrScheduler::step`: the internal step counter (usize) is incremented with `checked_add` on every scheduler step; after ~2^64 steps it would wrap, so the scheduler panics instead of silently corrupting milestone selection.

Source

Thrown at crates/burn-optim/src/lr_scheduler/sequential.rs:110

    schedulers: Vec<DynLrScheduler>,
    milestones: Vec<usize>,
    step: usize,
}

impl SequentialLrScheduler {
    fn active_scheduler(&self) -> usize {
        self.milestones.partition_point(|&m| self.step >= m)
    }
}

impl LrScheduler for SequentialLrScheduler {
    fn step(&mut self) -> LearningRate {
        let index = self.active_scheduler();
        let lr = self.schedulers[index].step();
        self.step = self
            .step
            .checked_add(1)
            .expect("The sequential scheduler step counter overflowed");
        lr
    }

    fn to_record(&self) -> LrSchedulerRecord {
        let mut record =
            LrSchedulerRecord::from_state(&SequentialLrSchedulerState { step: self.step });
        for (index, scheduler) in self.schedulers.iter().enumerate() {
            record = record.with_record(&index.to_string(), scheduler.to_record());
        }
        record
    }

    fn load_record(&mut self, record: LrSchedulerRecord) {
        if let Some(state) = record.into_state::<SequentialLrSchedulerState>() {
            self.step = state.step;
        }

        let schedulers = core::mem::take(&mut self.schedulers);

View on GitHub (pinned to d16f7ba2ed)

Solutions

1. Practically unreachable; only occurs with billions of scheduler steps
2. Reset the scheduler state via its record if running extremely long trainings
3. Reduce milestone/step frequency if step counts are artificially high

When it happens

Trigger: Thrown at crates/burn-optim/src/lr_scheduler/sequential.rs:110 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of tracel-ai/burn@d16f7ba2ed (2026-09-05). Data as JSON: /api/errors/d2982dee8b977493. Report an issue: GitHub.