trtllm-build Out of memory during weight-only INT4 quantization

The TensorRT-LLM engine build ran out of HOST memory, not GPU memory. Quantization holds full-precision weights in RAM while it converts them, so a 70B model can need well over 100 GB of system memory to build even though the resulting engine is small. This entry explains how to confirm the cause, apply the fix, and separate it from adjacent inference failures.

The literal signature is "trtllm-build Out of memory during weight-only INT4 quantization". It is a infrastructure failure associated with vLLM, Triton, and TensorRT-LLM servers. The line identifies the failing operation or subsystem, while the surrounding evidence decides whether it is the initiating fault or a downstream symptom.

The TensorRT-LLM engine build ran out of HOST memory, not GPU memory. Quantization holds full-precision weights in RAM while it converts them, so a 70B model can need well over 100 GB of system memory to build even though the resulting engine is small. 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 "trtllm-build Out of memory during weight-only INT4 quantization".
• 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.

• vLLM, Triton, and TensorRT-LLM servers
• 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 "trtllm-build Out of memory during weight-only INT4 quantization" and preserve at least 100 lines before it.
• ✓Identify which rank, node, device, or process emitted the first related warning.
• ✓the peak is roughly the full-precision checkpoint size plus the quantized copy. Confirm you are not building inside a container with a memory limit lower than the host's (docker run -m / Kubernetes limits), that is the most common cause of an OOM on a machine that appears to have plenty.
• ✓Repeat the same input after the targeted change and require the signature to disappear.
• ✓Resume from engine build restart on a larger host only after the control passes.

trtllm-build Out of memory during weight-only INT4 quantization

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.

build on a machine with more RAM, or add swap. Check the actual ceiling with free -g while the build runs, the failure is host-side. This changes the condition in the causal diagnosis instead of hiding the outer exception. The repeated control proves ownership before recovery from engine build restart on a larger host.

# 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 -- "trtllm-build Out of memory during weight-only INT4 quantization" <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 TensorRT-LLM engine build ran out of HOST memory, not GPU memory. Quantization holds full-precision weights in RAM while it converts them, so a 70B model can need well over 100 GB of system memory to build even though the resulting engine is small.
• The decisive evidence is the first log line that precedes "trtllm-build Out of memory during weight-only INT4 quantization" and differs from a healthy run.
• A nearby failure remains a competing hypothesis until a control separates configuration, capacity, transport, and hardware causes.