flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range

Quick answer

flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range means The FP8 attention scale or its reciprocal left the representable E4M3 range, so the kernel produced non-finite output. This is a scale-calibration failure before it is a general training-divergence failure. Preserve the first preceding error, then run the targeted control below.

Symptom: flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range
Root cause: The FP8 attention scale or its reciprocal left the representable E4M3 range, so the kernel produced non-finite output. This is a scale-calibration failure before it is a general training-divergence failure. The decisive evidence is the first log line that precedes "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range" and differs from a healthy run.
Recommended fix: reject the non-finite step, restore the last finite checkpoint, and run the batch in BF16 while recording the FP8 amax and scale values.
How Denpex helps: Denpex investigates flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range 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.

Training Stability#numerics#flash#attention#fp8#nan#matrix

What this failure is

The literal signature is "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range". It is a training stability failure associated with mixed precision, gradient scaling, and numerical kernels. The line identifies the failing operation or subsystem, while the surrounding evidence decides whether it is the initiating fault or a downstream symptom.

Live diagnosis, no signup

Is this what broke your run? Paste your log.

You're reading about flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range. 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:

3 free diagnoses/day

Why it happens (the mechanism)

The FP8 attention scale or its reciprocal left the representable E4M3 range, so the kernel produced non-finite output. This is a scale-calibration failure before it is a general training-divergence failure. The failure becomes visible at this call site because the operation first requires the missing resource, valid state, healthy peer, or correct result. Earlier log lines and a known-good control carry more causal value than the final wrapper exception.

What you'll observe

The workload stops or loses forward progress after emitting "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range".
A retry on the same configuration reproduces the failure because the causal state has not changed.
The outer framework exception can hide the rank, node, allocation, or dependency that failed first.
Increasing timeouts or reducing workload size can suppress the symptom without correcting the cause.

Common symptoms and what they mean

Symptom	Why it happens
flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range	The FP8 attention scale or its reciprocal left the representable E4M3 range, so the kernel produced non-finite output. This is a scale-calibration failure before it is a general training-divergence failure.
The same operation fails at a consistent stage of mixed precision, gradient scaling, and numerical kernels.	The decisive evidence is the first log line that precedes "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range" and differs from a healthy run.
The first related warning appears before the final exception and names the causal subsystem.	A nearby failure remains a competing hypothesis until a control separates configuration, capacity, transport, and hardware causes.
A known-good control changes one variable and either reproduces or clears the failure.	The FP8 attention scale or its reciprocal left the representable E4M3 range, so the kernel produced non-finite output. This is a scale-calibration failure before it is a general training-divergence failure.

Which systems are affected

mixed precision, gradient scaling, and numerical kernels
production-shaped multi-accelerator workloads
containerized and bare-metal deployments of the same stack

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.

✓Find the first occurrence of "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range" and preserve at least 100 lines before it.
✓Identify which rank, node, device, or process emitted the first related warning.
✓locate the first zero, infinite, or stale scale and verify the delayed-scaling history and update interval. Compare the same tensors with BF16 attention.
✓Repeat the same input after the targeted change and require the signature to disappear.
✓Resume from last checkpoint before the first FP8 non-finite result only after the control passes.

Root cause

The FP8 attention scale or its reciprocal left the representable E4M3 range, so the kernel produced non-finite output. This is a scale-calibration failure before it is a general training-divergence failure.
The decisive evidence is the first log line that precedes "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range" and differs from a healthy run.
A nearby failure remains a competing hypothesis until a control separates configuration, capacity, transport, and hardware causes.

The fix and how to prevent it

Searchable error signature

search key

flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range

Use this text as a lookup key in logs and upstream issue trackers. It is not presented as a captured customer log. Confirm the cause from your own preceding events, versions, configuration and the cited references.

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.

Why the recommended fix works

reject the non-finite step, restore the last finite checkpoint, and run the batch in BF16 while recording the FP8 amax and scale values. This changes the condition in the causal diagnosis instead of hiding the outer exception. The repeated control proves ownership before recovery from last checkpoint before the first FP8 non-finite result.

Code examples

snippet

# Preserve evidence before restarting
rg -n -i 'error|exception|timeout|failed' <log-file>
nvidia-smi
python -m torch.utils.collect_env

# Find the exact signature in the complete log
rg -n -F -- "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range" <log-file>

Adapt the snippet to your framework. The same pattern holds for PyTorch Lightning, Hugging Face Trainer, DeepSpeed, Megatron-LM, and vLLM training wrappers. Where the wrapper exposes a config flag (for examplelr_scheduler_type in Trainer), prefer the flag over the imperative API to keep the schedule declarative and reproducible.

Best practices by model family

Model / Stack	Recommendation	Notes
First response	Preserve the first failure	Keep the context before "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range" so aggregation does not erase causality.
Confirmation	Change one variable	Use a known-good node, rank, input, or configuration as the control.
Recovery	Resume from last checkpoint before the first FP8 non-finite result	Resume only after the literal signature no longer appears in the same control.

With the fix vs without the fix

Dimension	With the fix	Without the fix
Evidence	First preceding error and one controlled comparison	Only the final aggregated exception
Fix	reject the non-finite step, restore the last finite checkpoint, and run the batch in BF16 while recording the FP8 amax and scale values.	Retrying the unchanged workload
Exit criterion	"flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range" is absent in the repeated control	The job happened to run once

Diagnostic note

“Treat "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range" as a search key and an investigation checkpoint, not as proof of every cause associated with the phrase. The high-value evidence is what changed immediately before it and whether the failure follows the workload, node, or configuration.”

Visual fingerprint

Decision path for flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range

literal error captured
        |
        v
find first preceding failure
        |
        v
run one known-good control
        |
        +-- follows workload --> inspect input or configuration
        +-- follows node ------> inspect hardware or platform
        +-- disappears --------> validate the targeted fix

The control separates workload, configuration, and node ownership before recovery from last checkpoint before the first FP8 non-finite result.

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

Related failures to investigate next

Frequently asked questions

Questions engineers and on-call staff commonly ask about this failure.

What does "flash_attn_func produced all NaN values scale factor reciprocal overflowed FP8 E4M3 range" mean?

Is this line always the root cause?

No. It can be the direct failure or the point where an earlier failure becomes visible. The first preceding error and a controlled comparison decide which.

What should I collect before restarting?

Collect complete log context, the emitting rank or node, component versions, resolved configuration, and the diagnostic output shown above.

What is the fastest confirmation?

locate the first zero, infinite, or stale scale and verify the delayed-scaling history and update interval. Compare the same tensors with BF16 attention.

How do I prevent it from recurring?

clamp scale reciprocals to a safe finite range, alert on amax and scale saturation, and validate FP8 recipes per model and GPU generation.

References

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 Training Stability errors

NaN Detection and Skip

Training Stability · high

LR Without Warmup or Decay

Training Stability · high

Learning Rate Finder Result Misuse

Training Stability · medium

Gradient Accumulation BatchNorm Issue

Training Stability · high