Total 55,549 skills, AI & Machine Learning has 9240 skills
Showing 12 of 9240 skills
Used when an Agent needs to control OpenTeam via the local openteamcli: create AI group chats, add temporary roles, publish tasks, wait for replies, read results, or continue existing group chats.
Integrate TileGym kernels into Hugging Face `transformers` models by replacing the library's submodule(s) and certain class(es)' implementations, and patching certain class(es)' init/forward/load weight methods prior to instantiating models. Used when the user requires integrating TileGym kernels into `transformers` models.
Structured framework for verifying numerical parity of HF<->MCore weight conversions. References existing tools and the add-model-support skill.
Systematic workflow for MoE training optimization in Megatron Bridge, based on the Megatron-Core MoE paper. Covers the Three Walls framework, parallel folding, recompute strategy, dispatcher choice, and CUDA-graph bring-up.
Modify, build, test, debug, and contribute to NVIDIA cuOpt (C++/CUDA, Python, server, CI). Use for solver internals, PRs, DCO, and code conventions.
Use when debugging a Nemo Gym run or reward profiling job. Covers rollout collection failures, empty or partial JSONL outputs, stale materialized inputs, verifier/schema errors, Ray or Slurm issues, vLLM readiness, judge failures, tool/sandbox failures, cache problems, and throughput bottlenecks.
Generate video summary reports using the VSS video_search_frag extension with Long Video Summarization (LVS), Enterprise RAG knowledge retrieval, and human-in-the-loop parameter collection. Use when: user wants to generate a video summary, report, or analysis using the frag pipeline.
Code instrumentation for timing workloads. Two scenarios: (1) Training loop — inject manual timing to report per-iteration latency, throughput (samples/sec), and data load time. (2) Standalone kernel/op — write CUDA event timing code with warmup, per-iteration statistics, and anti-pattern avoidance. Also covers NVTX annotation for labeling profiler timelines. NOT for: running or analyzing profiler tools (nsys, ncu, Nsight Systems, Nsight Compute), writing kernels (Triton, CuTe, CUDA), applying optimizations (CUDA Graphs, gradient checkpointing, fusion), or interpreting roofline/SOL% metrics. Triggers: "measure throughput", "benchmark this function", "time my training loop", "samples per second", "NVTX annotate", "instrument my dataloader", "data load time", "kernel timing", "how do I time".
MoE expert-parallel communication overlap in Megatron Bridge. Covers dispatch/combine overlap, flex dispatcher backends, and expert wgrad scheduling.
Command-line interface for MiniMax AI — chat (MiniMax-M2.7) and TTS (speech-2.8-hd) via the MiniMax API.
Expert in deploying and using Hermes HUD Web UI for monitoring AI agent memory, sessions, costs, and health
Wang Jianshuo's thinking framework and expression style. Based on 7 in-depth researches (about 1 million words of English blogs + about 1.09 million words of Chinese blogs, 2002–2022, all first-hand), 7 core mental models, 10 decision-making heuristics and a complete bilingual expression DNA are extracted. Purpose: Write, respond and think from Wang Jianshuo's identity and voice—he is plain, sincere and curious, loves to use everyday metaphors and self-created words, repeatedly builds ladders between the concrete and the abstract, and never writes anything he hasn't personally verified. Activate when users mention phrases like "from Wang Jianshuo's perspective", "what would Wang Jianshuo think", "Wang Jianshuo's mode", "write like Wang Jianshuo", "Jian Shuo Wang perspective", "switch to Wang Jianshuo". It should also be triggered even if users only say "help me think from Wang Jianshuo's angle" or "how would Wang Jianshuo write this article". Once activated, all subsequent responses in this conversation will maintain Wang Jianshuo's identity until the user explicitly says "exit"—no need to name him repeatedly in each round. Inapplicable scenarios: When users ask for objective introductions or factual inquiries about Wang Jianshuo himself (such as "Who is Wang Jianshuo" or "When did he start his business"), answer such questions normally without entering role-playing.