AMD's skills help AI agents diagnose and repair ROCm, HIP, and PyTorch failures on AMD GPUs and serve LLMs with vLLM on EPYC CPUs and Instinct accelerators. They also cover local AI integration with Lemonade, Quark installation, PyTorch profiler trace analysis with TraceLens, and benchmarking and kernel optimization with Magpie and Hyperloom.
Skills
15 skills · 481 stars · MIT
Apu Memory Tuner Inspects and tunes the shared-vs-dedicated memory split on AMD Ryzen APUs with unified memory (UM…
Diagnose and fix ROCm, HIP and PyTorch failures on AMD GPUs Find out why PyTorch or llama.cpp can't see your AMD GPU, and fix it.
Eval Post Processing Aggregates TraceLens analysis repeatability CSVs into unit and e2e metric tables, classifies failure
Hyperloom Workload Optimizer Autonomously optimizes end-to-end LLM inference throughput on AMD Instinct GPUs and reports a val…
Lemonade Router Builder Turns a natural-language description of routing intent into a valid Lemonade `collection.router`…
Local Ai App Integration Integrates local AI capabilities into applications using Embeddable Lemonade.
Local Ai Use Makes this agent generate images, transcribe audio, and synthesize speech on the user's own machi…
Magpie Kernel Evaluator Benchmarks LLM inference and drives GPU kernel optimization with Magpie.
Quality Llm Eval Runs LLM-based quality evals 2 and 3 on generated analysis.md and category_findings files, comparing
Quark Install Installs or verifies AMD Quark and ensures the selected Python environment has an accelerator-mat…
Quark Torch Llm Ptq Runs an end-to-end AMD Quark post-training quantization workflow for PyTorch / Hugging Face LLMs:…
Serving Llms On Epyc Serves an LLM on a supported AMD EPYC server CPU using vLLM with zentorch, in Docker, Podman, or…
Serving Llms On Instinct Serves AI models on AMD Instinct GPU hardware using vLLM.
Tracelens Analysis Orchestrator Orchestrates modular PyTorch profiler trace analysis with TraceLens: generates perf reports, prep…
Workflow Llm Eval Runs LLM-based workflow eval 12 by checking the analysis.md Appendix hardware reference values (plat