Route AMD Quark user intents to the correct downstream skill or workflow.
npx skills add https://github.com/amd/Quark --skill quark-torch-router
Please help me install this Agent Skill. Skill: quark-torch-router Source: https://github.com/amd/Quark/tree/main/.claude/skills-impl/l1-atomic/torch/quark-torch-router Command: npx skills add https://github.com/amd/Quark --skill quark-torch-router
This Skill removes ambiguity when a user asks for help with AMD Quark by translating plain-language goals into the correct downstream skill or workflow. It prevents premature execution, wrong artifact selection, and unnecessary detours through the Quark skill stack.
Ask the quark-torch-router to interpret your Quark request and route it to the correct skill while documenting any missing environment or workspace facts.
High-intent search queries and answers about installing and using this skill.
AMD Quark model quantization requires routing through the full PTQ workflow to prevent incorrect execution. The router intercepts your request, validates prerequisites, and directs you to the correct multi-step workflow instead of direct script execution.
Quark quantization fails when bypassing the required PTQ workflow. The router enforces guardrails that prevent direct script execution, ensuring you follow the complete multi-step post-training quantization process for accurate results.
Yes, AMD Quark supports model inspection before quantization. The router detects your intent to inspect model architecture or metadata and directs you to the appropriate workflow step before proceeding to quantization planning.
Quark PTQ workflows require environment checks and workspace validation before execution. The router evaluates your session context, documents missing prerequisites, and validates environment and workspace artifacts to ensure successful quantization.
Debugging Quark quantization failures requires the router to identify the failure point within the PTQ workflow. It validates environment and workspace artifacts, documents missing prerequisites, and routes you to the correct debugging skill.
Yes, the Quark router maintains session context across all workflow steps. It carries forward references to your environment, workspace, installation, and workflow artifacts to ensure continuity throughout the PTQ routing process.
VetCodersBootstrap agent sessions with History, Eyes, and Verify context layers.
amdValidate workspace paths, model references, and output directories before Quark workflows.
amdRoutes Quark ONNX user requests to the smallest correct downstream workflow.
NVIDIAQuantize pretrained models to nvfp4, fp8, or int4_awq checkpoints using ModelOpt PTQ.
