Free browser engine · zero upload · no account
Offline GPU log analyzer
Run Denpex's deterministic offline diagnosis engine directly inside your browser. This tool analyzes CUDA OOMs, NCCL collective timeouts, NVIDIA Xid hardware faults, vLLM serving crashes, DeepSpeed errors, and Slurm cgroup kills into verifiable root-cause evidence and next commands with zero data upload.
Choose a synthetic sample or paste your own log text up to 512 KiB. All pattern matching, signature classification, and evidence extraction execute entirely in your local session.
Synthetic demonstration samples
Synthetic demonstration samples. Not customer telemetry.
Upload your log file. Large files are processed in your browser. Only selected evidence is submitted when you run diagnosis.
Processed entirely in your browser. Zero data leaves your machine.
Add host kernel / dmesg or nvidia-smi context (optional)
100% local browser engine (same deterministic core as the Open VSX & VS Code extension). 0 bytes uploaded.
What local browser diagnosis establishes
It identifies covered signature classes, separates initiating rank failures from secondary timeout wrappers, and generates actionable remediation commands and verification steps instantly in your browser.
Where multi-node fleet correlation takes over
Local analysis stops at a single buffer or file. Correlating cluster-wide clock skew, multi-node InfiniBand topology, and multi-tenant Slurm job queues requires Denpex's team evaluation and enterprise collector.
Frequently asked questions
Does any log data leave my browser on this page?
No. The entire diagnosis engine runs locally in your browser via WebAssembly and TypeScript with zero network egress and zero data upload.
Is this the same offline engine as the Open VSX and VS Code extension?
Yes. It uses the exact same deterministic canonical offline diagnosis core as our VS Code and Open VSX extensions.
Why does a multi-node NCCL timeout need logs from every rank?
A watchdog timeout only shows which rank waited too long. Pinpointing the root cause requires checking preceding errors from all available ranks.
What is the 512 KiB input cap and how are large logs handled?
Inputs are capped at 512 KiB to ensure fast browser performance. Large log files can be uploaded and intelligently sampled for final-tail and error context.
Ready to integrate offline diagnostics into your workflow?
Install our free extension or explore failure references.
Explore references: Xid, NCCL errors, CUDA errors, vLLM.