Is this a hardware or a software issue?

Model failures can look like either layer depending on the symptom. The discriminator is the resolution of the symptom to the recommended fix: if applying the steps under "How to fix it" eliminates the symptom, this is a software/configuration class issue regardless of the hardware it surfaced on. Repeat the fix on a fresh node to confirm.

How long does the fix typically take?

Most resolutions from this entry take 10 to 30 minutes of engineer time once the diagnosis is confirmed. Long fixes come from skipping the diagnosis step and chasing apparent symptoms (validation drift, memory growth, slow throughput) instead of the actual root cause listed above. Apply the fix in a small experiment first, then roll it forward.

Does this affect inference or only training?

Primarily this is a model failure observed during training, but the same misconfiguration can leak into downstream evaluation and serving. Treat the prevention checklist as a guardrail: any pipeline that touches a trained checkpoint should re-validate the failure mode before promoting the model.

What's the cheapest way to confirm this is the problem?

Cheapest signal: reproduce the symptom on a 1-GPU toy run with the same configuration. If the toy run exhibits the same behavior, the cause is the configuration rather than the cluster. If only the multi-GPU run is affected, the failure is a distributed interaction, so Megatron-LM is the most likely layer to inspect.

Can the Denpex platform detect this automatically?

Yes. Denpex correlates the symptoms in this entry against your training logs in real time and surfaces the failure taxonomy, root cause, and exact remediation command on the console. Three full diagnoses are included free during signup, and the encyclopedia entry you are reading is generated from the same signal library that powers the platform.

What related failures should I check next?

Closely related failures cluster around #Pipeline Configuration. After applying this fix, validate against the related entries below to confirm no overlapping cause is masking as this one. Engineers routinely discover that the dominant failure mode is actually a second, sibling failure compounded on top.

Will this recur after the fix?

Applied once with the prevention checklist, this failure does not recur under normal operation. The recurring variant most engineers encounter is the re-emergence of the same symptom on a different model or stack after Pipeline Configuration is upgraded. Pin compatible versions and rerun the prevention smoke test on every dependency change.

Is there a fingerprint I can monitor?

Yes. Training initialization succeeds, but hangs indefinitely upon completing the first forward pass or when logging auxiliary losses. is the highest-signal fingerprint. Wire a metric or alert on that signal (or its absence after the fix) so you can detect the recurrence automatically. The Denpex platform encodes this exact fingerprint and surfaces it directly in the dashboard.

What is the official documentation that covers this?

Reference documentation is listed in the References section at the bottom of this entry. The most authoritative URLs are the upstream framework docs and the issue trackers cited; the Denpex entry synthesizes both into the executive, engineer, and on-call perspectives so each audience can act without reading every linked source.

Does this entry include working code?

Yes. The "Code Examples" section contains copy-paste-ready snippets that target the most common stacks (PyTorch, Hugging Face Trainer, DeepSpeed, Megatron-LM, vLLM/Transformers). Each snippet is annotated with why it works so you can adapt the pattern to your own framework without losing fidelity.

MoE Pipeline Stage Asymmetry Hang

Q: What does this failure look like in production?

In production this failure typically appears during the first tens-to-hundreds of training steps: loss curves become non-monotonic, validation metrics plateau above the baseline, and downstream checkpoints either diverge or fail to improve. Engineers frequently misread it as data drift when the underlying mechanism is when using a custom `--pipeline-model-parallel-layout`, users may configure a pipeline stage to have zero moe layers (e.g., placing all moe layers in the middle stages). however, the logging/loss aggregation function `reduce_aux_losses_tracker_across_ranks` implicitly assumes all ranks in the pipeline group possess moe layers and invokes an allreduce across the entire pipeline group. the stage with 0 moe layers bypasses the moe loss logic and never joins the allreduce, causing the other stages to hang forever waiting for it.. Validating against the symptom list in this entry is the fastest way to confirm the diagnosis.

Q: How does this scale from 1 GPU to 1000 GPUs?

At single-GPU scale the symptom usually looks like slow convergence or a single anomalous loss value. At 64+ GPU scale, distributed interactions amplify when using a custom `--pipeline-model-parallel-layout`, users may configure a pipeline stage to have zero moe layers (e.g., placing all moe layers in the middle stages). however, the logging/loss aggregation function `reduce_aux_losses_tracker_across_ranks` implicitly assumes all ranks in the pipeline group possess moe layers and invokes an allreduce across the entire pipeline group. the stage with 0 moe layers bypasses the moe loss logic and never joins the allreduce, causing the other stages to hang forever waiting for it. through collectives, and the failure can manifest as a full job abort. Always reproduce the fix at the smallest cluster configuration that still exhibits the symptom before promoting to production scale.

When using a custom `--pipeline-model-parallel-layout`, users may configure a pipeline stage to have zero MoE layers (e.g., placing all MoE layers in the middle stages). However, the logging/loss aggregation function `reduce_aux_losses_tracker_across_ranks` implicitly assumes all ranks in the pipeline group possess MoE layers and invokes an AllReduce across the entire pipeline group. The stage with 0 MoE layers bypasses the MoE loss logic and never joins the AllReduce, causing the other stages to hang forever waiting for it.

Quick answer

When using a custom `--pipeline-model-parallel-layout`, users may configure a pipeline stage to have zero MoE layers (e.

Root cause: When using a custom `--pipeline-model-parallel-layout`, users may configure a pipeline stage to have zero MoE layers (e.g., placing all MoE layers in the middle stages).
Recommended fix: Balance MoE Layers or Patch Reduction Group Ensure each PP stage has at least 1 MoE layer in your layout. If all stages participate in the MoE logic, the AllReduce will succeed. Alternatively, patch the Megatron code to use a custom NCCL communicator that only includes ranks containing MoE layers.
How Denpex helps: Denpex investigates MoE Pipeline Stage Asymmetry Hang using the evidence you provide or your connected workload collects. Earlier rank, host or application evidence is needed to distinguish an initiating failure from a downstream report.

Model#Pipeline Configuration

What this failure is

MoE Pipeline Stage Asymmetry Hang is a Model failure seen during ML training runs. When using a custom `--pipeline-model-parallel-layout`, users may configure a pipeline stage to have zero MoE layers (e.g., placing all MoE layers in the middle stages). However, the logging/loss aggregation function `reduce_aux_losses_tracker_across_ranks` implicitly assumes all ranks in the pipeline group possess MoE layers and invokes an AllReduce across the entire pipeline group. The stage with 0 MoE layers bypasses the MoE loss logic and never joins the AllReduce, causing the other stages to hang forever waiting for it. Common tags: Pipeline Configuration.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about MoE Pipeline Stage Asymmetry Hang. Paste your own traceback and relevant evidence for an investigation of your workload, with a next action or a specific missing fact. A reference entry does not establish your cause. No account or card for the free diagnosis. Review data handling before submitting sensitive logs.

Before uploading, review cloud data handling and local options.

training_logs.txt

No log to hand? Try one:

Upload or drop a plain-text log

Upload your log file. Large files are processed in your browser. Only selected evidence is submitted when you run diagnosis.

3 free diagnoses/day

Why it happens (the mechanism)

Users assume the layout is valid because the model builds and allocates memory without errors. The failure only hits during the specific synchronization point for auxiliary losses.

What you'll observe

Process stuck in reduce_aux_losses_tracker_across_ranks
NCCL timeout waiting on rank in pipeline group

Common symptoms and what they mean

Symptom	Why it happens
Training initialization succeeds, but hangs indefinitely upon completing the first forward pass or when logging auxiliary losses.	When using a custom `--pipeline-model-parallel-layout`, users may configure a pipeline stage to have zero MoE layers (e.g., placing all MoE layers in the middle stages). However, the logging/loss aggregation function `reduce_aux_losses_tracker_across_ranks` implicitly assumes all ranks in the pipeline group possess MoE layers and invokes an AllReduce across the entire pipeline group. The stage with 0 MoE layers bypasses the MoE loss logic and never joins the AllReduce, causing the other stages to hang forever waiting for it.
Occurs specifically when using custom pipeline layouts with MoE models.	When using a custom `--pipeline-model-parallel-layout`, users may configure a pipeline stage to have zero MoE layers (e.g., placing all MoE layers in the middle stages). However, the logging/loss aggregation function `reduce_aux_losses_tracker_across_ranks` implicitly assumes all ranks in the pipeline group possess MoE layers and invokes an AllReduce across the entire pipeline group. The stage with 0 MoE layers bypasses the MoE loss logic and never joins the AllReduce, causing the other stages to hang forever waiting for it.

Symptom

Why it happens

Training initialization succeeds, but hangs indefinitely upon completing the first forward pass or when logging auxiliary losses.

Occurs specifically when using custom pipeline layouts with MoE models.

Which systems are affected

Megatron-LM
Pipeline Parallelism

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.

✓Inspect the `--pipeline-model-parallel-layout` argument and cross-reference with where MoE layers are injected.
✓Use `py-spy` and check if ranks are blocked inside `reduce_aux_losses_tracker_across_ranks`.
✓Check if `--moe-router-load-balancing-type` implies aux loss calculation.

Root cause

When using a custom `--pipeline-model-parallel-layout`, users may configure a pipeline stage to have zero MoE layers (e.g., placing all MoE layers in the middle stages). However, the logging/loss aggregation function `reduce_aux_losses_tracker_across_ranks` implicitly assumes all ranks in the pipeline group possess MoE layers and invokes an AllReduce across the entire pipeline group. The stage with 0 MoE layers bypasses the MoE loss logic and never joins the AllReduce, causing the other stages to hang forever waiting for it.

The fix and how to prevent it

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.

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

References

↗GitHub Issue #1852: Custom Pipeline Layout Deadlocks

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.

Diagnose your logs free Compare a solved incident Browse all failures

Related Model errors

FSDP Hang Due to Divergent Parameter Initialization

Model · high

RMSNorm Variance Overflow in FP16

Model · high

Learned Temperature Collapse in Contrastive Loss

Model · high

Attention Softmax FP16 Overflow

Model · high