https://errorvault.dev/tags/clw-llm-crash/Recent content in Clw-Llm-Crash on ErrorVault — Developer Error Code DictionaryHugoen-usMon, 13 Apr 2026 22:24:49 +0800https://errorvault.dev/openclaw/openclaw-clw-llm-crash-inference-failure/Mon, 13 Apr 2026 22:24:49 +0800https://errorvault.dev/openclaw/openclaw-clw-llm-crash-inference-failure/<h2 id="1-symptoms">1. Symptoms</h2> <p>The <code>clw-llm-crash</code> error in OpenClaw manifests as an abrupt termination of the LLM inference process. Common indicators include:</p> <ul> <li><strong>Core dump or segmentation fault</strong>: Process exits with signal 11 (SIGSEGV) during model loading or token generation.</li> <li><strong>Log output</strong>: <div class="highlight"><pre tabindex="0" style="color:#f8f8f2;background-color:#282a36;-moz-tab-size:4;-o-tab-size:4;tab-size:4;"><code class="language-fallback" data-lang="fallback"><span style="display:flex;"><span>[ERROR] clw-llm-crash: LLM backend failed at tensor allocation (line 1423, llm_engine.cpp) </span></span><span style="display:flex;"><span>[FATAL] GPU context lost: CUDA error 700 (cudaErrorIllegalAddress) </span></span><span style="display:flex;"><span>Aborted (core dumped) </span></span></code></pre></div></li> <li><strong>High resource usage spike</strong>: VRAM usage jumps to 100% followed by OOM killer activation or driver reset.</li> <li><strong>Reproducible on specific models</strong>: Crashes consistently with quantized GGUF models (e.g., Llama-3-8B-Q4_K_M.gguf) but not unquantized ones.</li> <li><strong>Platform-specific</strong>: More frequent on NVIDIA GPUs with CUDA 11.x; AMD ROCm users report HIP kernel panics.</li> </ul> <p>Stack traces often point to <code>llm_engine.cpp</code> or <code>claw_cuda_backend.cu</code> in OpenClaw source. Use <code>gdb</code> or <code>cuda-gdb</code> for deeper inspection:</p>