Skip to content

Memory Fragmentation

Memory fragmentation causes OOM even when total free memory is sufficient, because no contiguous block is available for the allocation.

Quick answer

Memory fragmentation causes OOM even when total free memory is sufficient, because no contiguous block is available for the allocation.

Memory#cuda#memory","fragmentation","pytorch","oom#allocator

What this failure is

Memory Fragmentation is a Memory failure seen during ML training runs. Memory fragmentation causes OOM even when total free memory is sufficient, because no contiguous block is available for the allocation. Common tags: Cuda, Memory","Fragmentation","Pytorch","Oom, Allocator.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about Memory Fragmentation. Paste your own crash log or traceback below and get the real root cause for YOUR run, not this generic entry. No account, no card. Logs are masked at ingress and never saved to account history.

training_logs.txt
No log to hand? Try one:

3 free diagnoses/day

Want 14 days on the Scale plan?

Request an evaluation code. A verified workplace organization activates up to 50 diagnoses a day, alerts, history, and follow-up questions. No credit card or automatic subscription.

Evaluate one incident

Why it happens (the mechanism)

Variable-size allocations create non-contiguous free blocks. PyTorch's caching allocator doesn't merge free blocks. Long-running training accumulates fragmentation over thousands of steps. 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.

What you'll observe

  • OOM with seemingly sufficient free memory
  • Intermittent OOM that sometimes works on retry
  • OOM after many training steps but not at start

Common symptoms and what they mean

SymptomWhy it happens
torch.cuda.memory_summary() shows fragmented free blocksVariable-size allocations create non-contiguous free blocks
OOM for allocation size that fits in total free memoryPyTorch's caching allocator doesn't merge free blocks
Different OOM behavior across runs with same configurationLong-running training accumulates fragmentation over thousands of steps

Which systems are affected

  • Long-running training with varying tensor sizes
  • Training with dynamic batch sizes
  • Models with attention layers of varying sequence length

How to confirm this is the problem

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: torch.cuda.memory_summary() shows fragmented free blocks
  • Verified signal present: OOM for allocation size that fits in total free memory
  • Verified signal present: Different OOM behavior across runs with same configuration
  • A targeted fix from the "How to fix it" section eliminates or substantially reduces the symptom within one validation pass.

The fix and the prevention pattern

The root cause is on this page and stays free. A free account adds the exact remediation steps, saved history, and the fix on every entry in the encyclopedia.

Sign up free. Unlock the full analysis

No credit card. Daily allowance follows verified trust tier. Instant access.

Diagnose this failure in VS Code

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.

Install the free VS Code extension

CUDA errors in context

CUDA reports errors asynchronously, so the traceback usually points at whatever line synchronised next rather than the one at fault. The hub covers every common CUDA error and how to make it report honestly.

Compare every cuda error side by side

Root cause

  • Variable-size allocations create non-contiguous free blocks
  • PyTorch's caching allocator doesn't merge free blocks
  • Long-running training accumulates fragmentation over thousands of steps

The fix and how to prevent it

Evaluate Denpex on your own logs

Request a Scale evaluation code. A verified workplace organization activates 14 days with up to 50 diagnoses a day. Every account keeps its current diagnosis allowance and gets a verification path. No card or automatic subscription.

We send a single-use code tied to that address. Static provider and TLD rules do not reject valid addresses. Account trust determines the benefit after signup.

Don't just read the fix, diagnose your run

The encyclopedia tells you what went wrong. Denpex tells you what went wrong in YOUR training run. With your logs, your config, and your stack.