AMP BF16 vs FP16 Confusion
BF16 and FP16 mixed precision have different numerical properties; choosing the wrong one causes instability or wasted memory.
BF16 and FP16 mixed precision have different numerical properties.
What this failure is
AMP BF16 vs FP16 Confusion is a Training Stability failure seen during ML training runs. BF16 and FP16 mixed precision have different numerical properties; choosing the wrong one causes instability or wasted memory. Common tags: Bf16, Fp16, Mixed Precision, Gradscaler.
Is this what broke your run? Paste your log.
You're reading about AMP BF16 vs FP16 Confusion. 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.
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.
Why it happens (the mechanism)
FP16 has limited exponent range causing overflow. BF16 has same exponent as FP32 but less precision. FP16 needs loss scaling, BF16 doesn't. BF16 is preferred for H100/A100. 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
- Loss is unstable with FP16
- FP16 overflow causes NaN
- BF16 uses more memory than expected
Common symptoms and what they mean
| Symptom | Why it happens |
|---|---|
| FP16 NaN with certain operations | FP16 has limited exponent range causing overflow |
| BF16 model is less accurate than FP32 | BF16 has same exponent as FP32 but less precision |
| Mixed precision doesn't speed up training | FP16 needs loss scaling, BF16 doesn't |
Which systems are affected
- Training with torch.cuda.amp
- Mixed precision training
- Using H100/A100 with BF16
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: FP16 NaN with certain operations
- ✓Verified signal present: BF16 model is less accurate than FP32
- ✓Verified signal present: Mixed precision doesn't speed up training
- ✓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 analysisNo 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 extensionRelated failures to investigate next
Root cause
- FP16 has limited exponent range causing overflow
- BF16 has same exponent as FP32 but less precision
- FP16 needs loss scaling, BF16 doesn't
- BF16 is preferred for H100/A100
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.
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.