loom

Plugin loom from marketplace loom in loom-fun/loom-skills. Commands read from the repo's manifest — yesterday.

Skills (9)

• loom-benchmark

  Correctness and performance methodology for Loom models vs llama.cpp ROCm — golden token-by-token comparison with log-prob near-tie analysis, a multi-turn tool-calling (agent) gate, a full-context needle gate, batching consistency, decode/prefill/long-context/concurrency throughput on real-text work

• loom-demo-ui

  Start and use the Loom Playground web UI (chat with streaming reasoning and tok/s, ASR with mic recording, TTS, image generation, embedding similarity) that proxies to one or more running loom-serve instances and discovers their models automatically. Use when the user wants a browser demo/UI for Loo

• loom-improve

  Improves the loom skills themselves (port, optimize, qualify, report) without regressing them - the KPIs (end-to-end wall clock, correctness vs llama.cpp in agentic use, decode/prefill speed), how to measure them with fresh-agent trials on a reference model suite, the improve-retest loop, subagent h

• loom-optimize

  Make a Loom LLM port faster (decode, prefill, long context, concurrency) while proving it stays correct — the measure/ablate/change/verify loop, correctness gates (kernel A/B tests, llama.cpp golden, concurrency consistency), and a catalog of optimizations that worked (and ones that didn't) on RDNA3

• loom-org

  Populate and maintain a Hugging Face org of Loom models (loom-fun) end to end - pick the top models that fit Strix Halo, classify what the org already has (category A current / B older skill / missing), then loop - port and optimize each model with a subagent on the local GPU starting from the best

• loom-port-llm

  Port GGUF LLMs to AMD Loom and loom-serve: inspect metadata, generate and test kernels, derive chat/tool/reasoning glue, build at native context, and validate against llama.cpp. Covers dense and MoE transformers, gated delta-net hybrids, speculative decoding and supported vision projectors, with a p

• loom-roofline

  Roofline analysis for a Loom LLM port on Strix Halo — compute the hardware ceiling for decode, long-context decode, prefill and 4-stream throughput from the GGUF, compare llama.cpp ROCm's efficiency, and pick realistic, achievable, impressive per-metric optimization targets with a stopping rule. Use

• loom-setup

  Install AMD Loom (hrx-loom from PyPI, or a ROCm/hrx-system CI wheel) and the hrx-loom-library catalog from a cold start, fetch catalog models with loom-fetch, and start/stop/inspect loom-serve on the GPU or NPU. Use for "install loom", "loom-fetch a model", "start loom-serve for X", "is the server r

• loom

  Entry point for AMD Loom work (hrx-loom SDK, loom-serve, loom-fetch, Loom IR kernels) on AMD GPUs/NPUs such as Strix Halo — installing, serving catalog models, the demo web UI, porting a new LLM to Loom, optimizing it while preserving correctness, and benchmarking against llama.cpp. Use whenever the

Similar plugins