OOM Killed Mid-Step

OOM kills during training are abrupt and lose progress, often happening mid-step when peak memory is reached.

OOM Killed Mid-Step is a Reliability failure seen during ML training runs. OOM kills during training are abrupt and lose progress, often happening mid-step when peak memory is reached. Common tags: Oom, Oom Killer, Memory, Reliability.

Peak memory exceeds available. Memory leak grows until OOM. Cgroup limit lower than job memory. Other processes consume memory. Taken together, these mechanisms explain why the failure is reproducible, why it tends to surface on specific workloads or scales, and why generic mitigation attempts often fall short without addressing the underlying cause.

• Training process is killed with OOM
• Job disappears with exit code 137
• No graceful shutdown on OOM

• Memory-intensive training jobs
• Long-running training with growing memory
• Cgroup-based resource limits

Use this checklist to test the hypothesis against a small reproduction. No single line proves the root cause, so preserve the preceding events and compare one variable at a time.

• ✓Reproduce the failure from a clean checkpoint/seed: the symptom must appear without warm-up state from a previous run.
• ✓Verified signal present: dmesg shows OOM killer
• ✓Verified signal present: Exit code 137 (SIGKILL)
• ✓Verified signal present: Process vanishes during training step
• ✓A targeted fix from the "How to fix it" section eliminates or substantially reduces the symptom within one validation pass.

Select the traceback or open the failed terminal, then run Denpex locally to see the initiating rank, collateral failures, exact fix, and verification command without uploading the log.

• Peak memory exceeds available
• Memory leak grows until OOM
• Cgroup limit lower than job memory
• Other processes consume memory