torchao quantization CalibrationError observed activation tensor contains NaN during INT8 observer

The INT8 observer saw NaN or Inf activations while calibrating. Quantization scales are derived from observed ranges, so a single non-finite activation poisons the scale for that tensor and every value quantized with it. This entry explains how to confirm the cause, apply the fix, and separate it from adjacent quantization failures.

The literal signature is "torchao quantization CalibrationError observed activation tensor contains NaN during INT8 observer". It is a environment failure associated with AWQ, GPTQ, bitsandbytes, GGUF, and torchao. The line identifies the failing operation or subsystem, while the surrounding evidence decides whether it is the initiating fault or a downstream symptom.

The INT8 observer saw NaN or Inf activations while calibrating. Quantization scales are derived from observed ranges, so a single non-finite activation poisons the scale for that tensor and every value quantized with it. 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.

• The workload stops or loses forward progress after emitting "torchao quantization CalibrationError observed activation tensor contains NaN during INT8 observer".
• 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.

• AWQ, GPTQ, bitsandbytes, GGUF, and torchao
• production-shaped multi-accelerator workloads
• containerized and bare-metal deployments of the same stack

Apply the following checklist to a small reproduction: each box below is a positive signal that you are looking at this exact failure rather than a sibling in the same taxonomy.

• ✓Find the first occurrence of "torchao quantization CalibrationError observed activation tensor contains NaN during INT8 observer" and preserve at least 100 lines before it.
• ✓Identify which rank, node, device, or process emitted the first related warning.
• ✓the usual sources are a calibration batch containing padding that the attention mask does not cover, or a model already unstable in fp16. Try a different calibration batch, if only one batch fails, the data is the problem.
• ✓Repeat the same input after the targeted change and require the signature to disappear.
• ✓Resume from restart calibration with a validated batch only after the control passes.

torchao quantization CalibrationError observed activation tensor contains NaN during INT8 observer

Timestamps and exact values vary across runs, but the pattern. An info-level start, an early WARN, an ERROR carrying the symptom. Is the actual fingerprint you should alert on. The Denpex platform flags this combination automatically.

find where the NaN originates, it is in the MODEL, not the quantizer. Run the calibration batch in full precision and assert torch.isfinite on activations layer by layer. This changes the condition in the causal diagnosis instead of hiding the outer exception. The repeated control proves ownership before recovery from restart calibration with a validated batch.

# 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 -- "torchao quantization CalibrationError observed activation tensor contains NaN during INT8 observer" <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.

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.

• The INT8 observer saw NaN or Inf activations while calibrating. Quantization scales are derived from observed ranges, so a single non-finite activation poisons the scale for that tensor and every value quantized with it.
• The decisive evidence is the first log line that precedes "torchao quantization CalibrationError observed activation tensor contains NaN during INT8 observer" and differs from a healthy run.
• A nearby failure remains a competing hypothesis until a control separates configuration, capacity, transport, and hardware causes.