ModelTCModelTC/LightX2V/tools/AGENTS.md

LightX2V / tools

AGENTS.md3k starsChanged 4 days agoApache-2.0View on GitHub

进行性能分析、benchmark 或优化前，先从以下正式入口查找可复用工具，阅读相关目录的 AGENTS 和 README，并按当前实现核对输入、输出与支持边界。沿目标模型的 scripts/、配置和真实调用链确定接入位置。按任务流程组合已有能力，明确需要补的模型接线及验证；工具缺少能力、需要模型接线和已有能力尚未验证分别处理。任务特有的输入捕获、重建与状态恢复放在任务脚本；稳定、可复用的精确操作按职责补入工具。源码足以判断或已有证据仍适用时复用，不重复运行完整测试矩阵。交付时区分工具产生的事实、Agent 补写的接线与解释、实际执行的验证范围。任务条件、进程、命令和产物保存到当前工作包；不改变已有资源权限或质量取舍。

Create a free account to mount

Files

AGENTS.md

性能工具导航

进行性能分析、benchmark 或优化前，先从以下正式入口查找可复用工具，阅读相关目录的 AGENTS 和 README，并按当前实现核对输入、输出与支持边界。

任务	入口
真实函数/模块采集、计算量关联与性能分析	Profiler 指南 · 接口与指标
Benchmark 与计算量	Benchmark 指南 · 接口与指标
GPU 资源观测	运行指南 · 工具用法

沿目标模型的 scripts/、配置和真实调用链确定接入位置。按任务流程组合已有能力，明确需要补的模型接线及验证；工具缺少能力、需要模型接线和已有能力尚未验证分别处理。

任务特有的输入捕获、重建与状态恢复放在任务脚本；稳定、可复用的精确操作按职责补入工具。源码足以判断或已有证据仍适用时复用，不重复运行完整测试矩阵。

交付时区分工具产生的事实、Agent 补写的接线与解释、实际执行的验证范围。任务条件、进程、命令和产物保存到当前工作包；不改变已有资源权限或质量取舍。

Download now (ZIP)

Oct 20, 2026Latest 15 Papers - October 20, 2026 — guoyue0412/DailyArxiv0 reactions
Oct 19, 2026Latest 10 Papers - October 20, 2026 — luohongk/Embodied-AI-Daily0 reactions
Oct 19, 2026Latest 20 Papers - October 20, 2026 — xllm-io/daily-papers0 reactions
Oct 19, 2026Latest 15 Papers - October 19, 2026 — guoyue0412/DailyArxiv0 reactions
Oct 19, 2026Latest 15 Papers - October 19, 2026 — Ed1sonChen/DailyArxiv0 reactions

20 more on GitHub

Oct 7, 2026Latest 20 Papers - October 08, 2026 — xllm-io/daily-papers0 reactions
Oct 7, 2026Latest 15 Papers - October 07, 2026 — guoyue0412/DailyArxiv0 reactions
Oct 7, 2026Latest 15 Papers - October 07, 2026 — Ed1sonChen/DailyArxiv0 reactions
Oct 6, 2026Latest 10 Papers - October 07, 2026 — luohongk/Embodied-AI-Daily0 reactions
Oct 6, 2026Latest 20 Papers - October 07, 2026 — xllm-io/daily-papers0 reactions
Oct 6, 2026Latest 15 Papers - October 06, 2026 — guoyue0412/DailyArxiv0 reactions
Oct 6, 2026Latest 15 Papers - October 06, 2026 — Ed1sonChen/DailyArxiv0 reactions
Sep 28, 2026[til] Qwen-Image-2.1 Review: Unified Text-to-Image Generation and Editing — sameerkhansf/Sameer-Khan0 reactions · 2 comments
Jul 14, 2026ModuleNotFoundError: No module named 'lightx2v_kernel' when launching Web UI with the latest public LightX2V — amap-cvlab/ABot-World0 reactions · 1 comment
May 26, 2026能否考虑合作一下，给LightX2V框架，支持上昆仑芯的后端 — baidu/vLLM-Kunlun0 reactions · 2 comments
Feb 10, 2026[Diffusion] opt vae decode with `channels_last_3d` — sgl-project/sglang0 reactions · 5 comments
Jan 20, 2026[Feature] Support DiT Layerwise (Blockwise) CPU Offloading — vllm-project/vllm-omni0 reactions · 7 comments
Jan 5, 2026Update doc to add integrated projects — MooreThreads/torchada0 reactions
Jan 4, 2026Serving: Align inference time format with LightX2V (timezone-aware) — vipshop/cache-dit0 reactions
Dec 24, 2025LightX2V: Light Video Generation Inference Framework, ModelTC, 2025.12 — AkihikoWatanabe/paper_notes0 reactions · 1 comment
Dec 22, 2025feat: Integrate Wan2.2 VAE, LightX2V, and TAE with VRAM optimization and Pre-Flight Resource Calculator — naxci1/ComfyUI-FlashVSR_Stable0 reactions · 14 comments
Nov 23, 2025Add pre-quantized checkpoint and super resolution support for HunyuanVideo-1.5 — Ok2dot0/ComfyUI-Lightx2vWrapper0 reactions
Nov 23, 2025Add support for HunyuanVideo-1.5 models — Ok2dot0/ComfyUI-Lightx2vWrapper0 reactions · 2 comments
Oct 30, 2025[Bug] compilation: Call to `scikit_build_core.build.build_wheel` failed — sgl-project/sglang0 reactions · 3 comments
Oct 30, 2025[Bug]: subprocess.CalledProcessError: Command '['cmake', '--build', '.', '-j=15', '--target=_moe_C', '--target=_vllm_fa2_C', '--target=_vllm_fa3_C', '--target=_flashmla_C', '--target=cumem_allocator', '--target=_C']' returned non-zero exit status 137. — vllm-project/vllm0 reactions · 2 comments

In this repository11 other files it gives its agents.
- benchmark-core-opsSkill
- benchmark-module-parallelismSkill
- model-enable-intel-xpuSkill
- profile-model-moduleSkill
- remove-ai-code-smellSkill
- support-contiguous-block-offloadSkill
- support-cpu-block-offloadSkill
- support-model-compileSkill
- support-model-warmupSkill
- support-new-modelSkill
- support-static-block-residencySkill

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free