Method _watchdog_loop

torch/distributed/elastic/timer/api.py:179–184

(self)

            return True

    def _watchdog_loop(self):
        while not self._stop_signaled:
            try:
                self._run_watchdog()
            except Exception:
                log.exception("Error running watchdog")

    def _run_watchdog(self):
        batch_size = max(1, self._request_queue.size())