{"version":"1.0","workflow_uuid":"b2a92e6b-453a-11f1-9bc6-00163e2b0d79","workflow_title":"PowerInfer — High-Speed Local LLM Inference via Activation Locality","install_contract":{"version":"1.0","installReady":false,"title":"PowerInfer — High-Speed Local LLM Inference via Activation Locality","summary":"A CPU/GPU LLM inference engine that exploits activation locality to achieve high-speed generation on consumer hardware. Runs large models efficiently by only computing activated neurons.","assetType":"Scripts","pageUrl":"https://tokrepo.com/en/workflows/powerinfer-high-speed-local-llm-inference-via-activation-b2a92e6b","sourceUrl":"https://github.com/SJTU-IPADS/PowerInfer","intendedFor":[],"firstActions":[],"agentFirstSteps":[],"targetPaths":[],"verification":[],"startingPoints":[],"example":"","successOutcome":"","boundaries":[],"askUserIf":["the current workspace stack cannot be matched to a safe upstream template","the target path is not the project root, or an existing file should be merged instead of overwritten"]}}