flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill

A FlashInfer paged-prefill kernel accessed an invalid GPU address. Common owners are malformed page metadata, a shape or dtype unsupported by the selected kernel, or a binary built for an incompatible CUDA stack. This entry explains how to confirm the cause, apply the fix, and separate it from adjacent fp8-serving failures.

Symptom

flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill

Root cause

A FlashInfer paged-prefill kernel accessed an invalid GPU address. Common owners are malformed page metadata, a shape or dtype unsupported by the selected kernel, or a binary built for an incompatible CUDA stack. The decisive evidence is the first log line that precedes "flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill" and differs from a healthy run.

Recommended fix

CUDA_LAUNCH_BLOCKING=1

How Denpex helps

Denpex investigates flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill 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.

The literal signature is "flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill". It is a infrastructure failure associated with FP8 serving with SGLang, FlashInfer, and vLLM. The line identifies the failing operation or subsystem, while the surrounding evidence decides whether it is the initiating fault or a downstream symptom.

A FlashInfer paged-prefill kernel accessed an invalid GPU address. Common owners are malformed page metadata, a shape or dtype unsupported by the selected kernel, or a binary built for an incompatible CUDA stack. 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 "flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill".
• 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.

• FP8 serving with SGLang, FlashInfer, and vLLM
• production-shaped multi-accelerator workloads
• containerized and bare-metal deployments of the same stack

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 "flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill" and preserve at least 100 lines before it.
• ✓Identify which rank, node, device, or process emitted the first related warning.
• ✓validate page indices, sequence lengths, dtype, head dimensions, and FlashInfer/PyTorch/CUDA compatibility. Run the same input with the fallback attention backend.
• ✓Repeat the same input after the targeted change and require the signature to disappear.
• ✓Resume from request retry in a fresh process after the kernel control passes only after the control passes.

• A FlashInfer paged-prefill kernel accessed an invalid GPU address. Common owners are malformed page metadata, a shape or dtype unsupported by the selected kernel, or a binary built for an incompatible CUDA stack.
• The decisive evidence is the first log line that precedes "flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill" and differs from a healthy run.
• A nearby failure remains a competing hypothesis until a control separates configuration, capacity, transport, and hardware causes.

flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill

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.

restart the process to clear the poisoned CUDA context, then reproduce one request with CUDA_LAUNCH_BLOCKING=1 and the smallest failing page table. This changes the condition in the causal diagnosis instead of hiding the outer exception. The repeated control proves ownership before recovery from request retry in a fresh process after the kernel control passes.

# 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 -- "flashinfer CUDA kernel launch failure cudaErrorIllegalAddress paged_prefill" <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.

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.