Loading...
Loading...
Found 12 Skills
Complete the full loop of "Data → Fine-tuning Training → Export → Deployment → Inference" using Bailian CLI (`bl`), or deploy base models directly without training. Supports fine-tuning of text models (SFT/DPO/CPT), audio TTS models (CosyVoice), and image generation models (Wan2.7). Covers dataset validation/upload, creating fine-tuning tasks, waiting for training completion, exporting the best checkpoint, creating inference deployments, waiting for readiness, and providing inference examples. This skill should be activated when users mention actions like "training models", "fine-tuning", "fine-tune", "finetune", "deploying models", "model launch", "running/calling fine-tuned models", "training an inference model", "continuing pre-training", "LoRA/SFT/DPO training", "speech synthesis models", "TTS fine-tuning", "CosyVoice", "voice cloning", "image generation fine-tuning", "text-to-image", "image-to-image", "Wan2.7", "image model training" on Bailian / DashScope / Alibaba Cloud Model Studio — even if users don't explicitly mention "using bl", as long as the intention is training or deployment on the Bailian platform, use this skill and do not assemble commands on your own.
Generates and updates secure, production-ready Kubernetes YAML manifests optimized for GKE Autopilot and GKE Standard clusters. Use when creating or modifying GKE deployment manifests, configuring container security contexts, setting CPU/memory resource limits, defining readiness/liveness/startup probes, mounting secrets and volumes, configuring GKE Gateway API routes, targeting Spot VMs, or deploying AI model inference workloads (vLLM, TGI, Gemma). Don't use for live cluster operations, pod troubleshooting (use gke-workload-troubleshooting), or cluster infrastructure provisioning (use gke-cluster-creation).
Connect to a running ComfyUI instance, queue workflows, monitor execution, and retrieve results. Supports both online (REST API) and offline (JSON export) modes. Use when executing ComfyUI workflows or checking server status.
Load for any work involving Baseten - deploying/operating models on Dedicated Inference (Truss, custom Docker servers, TRT-LLM engines, Chains), calling hosted Model APIs, running Training jobs (SFT/RL/LoRA), or Model Frontier Gateway.
Build and publish a Gradio demo on Hugging Face Spaces for a user-provided LoRA. Use when someone asks to create, generate, ship, or publish a Space, demo, Gradio app, or playground for a LoRA — including LoRAs for Qwen-Image, Qwen-Image-Edit, LTX-Video, Wan, FLUX, SDXL, or other diffusion base models. Also triggers when someone describes a LoRA they trained or hosts on the Hub and wants to share it. Covers picking the right base pipeline and `diffusers` inference recipe, designing a UI tailored to the LoRA's task and inputs (Union/multi-task control, edit, video, image, etc.), respecting model-card recommendations (trigger words, steps, guidance, LoRA scale, example inputs), and shipping to ZeroGPU hardware as a private Space by default.
Complete knowledge domain for Cloudflare Workers AI - Run AI models on serverless GPUs across Cloudflare's global network. Use when: implementing AI inference on Workers, running LLM models, generating text/images with AI, configuring Workers AI bindings, implementing AI streaming, using AI Gateway, integrating with embeddings/RAG systems, or encountering "AI_ERROR", rate limit errors, model not found, token limit exceeded, or neurons exceeded errors. Keywords: workers ai, cloudflare ai, ai bindings, llm workers, @cf/meta/llama, workers ai models, ai inference, cloudflare llm, ai streaming, text generation ai, ai embeddings, image generation ai, workers ai rag, ai gateway, llama workers, flux image generation, stable diffusion workers, vision models ai, ai chat completion, AI_ERROR, rate limit ai, model not found, token limit exceeded, neurons exceeded, ai quota exceeded, streaming failed, model unavailable, workers ai hono, ai gateway workers, vercel ai sdk workers, openai compatible workers, workers ai vectorize
Compact SGLang torch-profiler triage skill. Use when Codex should inspect an existing `trace.json(.gz)` or profile directory, trigger `sglang.profiler` against a live server, and return one compact report with kernel, overlap-opportunity, and fuse-pattern tables. Single-trace triage is enough for quick diagnosis; mapping+formal two-trace triage gives stronger overlap conclusions.
Inspect the availability of model serving on a completed Itô compute booking and, when the canonical backend becomes available, hand off an explicitly confirmed serving manifest. Use after ito-compute has booked GPU nodes and the user asks for an OpenAI-compatible endpoint, ito-serve, hosted Kimi, or self-hosted open-weights inference. ECC implements no serving stack of its own.
High-level map of the Venice.ai API - base URL, authentication modes, endpoint categories, response headers, pricing model, error shape, and versioning. Load this first when starting any Venice integration.
User-facing NemoClaw guidance for installing, configuring, operating, securing, monitoring, and troubleshooting NemoClaw sandboxes. Use when users ask about NemoClaw quickstarts, OpenClaw and OpenShell relationships, local inference, remote GPU deployment, sandbox lifecycle, network policy, security posture, agent skills, command reference, or issue triage instructions.
Compile TensorRT-LLM on a compute node inside a Docker container. Use this when already on a compute node with GPUs visible.
Replay-first debug flow for SGLang serving problems. Use when a live or recent server shows health-check failures, latency or throughput regressions, queue growth, timeouts, distributed stalls, crash dumps, wrong outputs after deploys, or PD/EP/HiCache issues, and the job is to turn the problem into a replay plus the right next debug tool.