Provides ms-swift CLI guidance for training, RLHF, quantization, and OpenAI-compatible deployment workflows.
npx skills add https://github.com/wenerme/ai --skill swift-ms-docs
Please help me install this Agent Skill. Skill: swift-ms-docs Source: https://github.com/wenerme/ai/tree/main/skills/swift-ms-docs Command: npx skills add https://github.com/wenerme/ai --skill swift-ms-docs
This Skill solves the problem of quickly finding the right ms-swift (ModelScope Swift) commands, concepts, and reference docs needed to fine-tune, run reinforcement learning, and deploy models without getting lost in scattered guidance.
Use the swift-ms-docs skill when you need to get ms-swift fine-tuning or GRPO command-line instructions immediately for your target model and dataset.
High-intent search queries and answers about installing and using this skill.
ms-swift fine-tuning supports LoRA, QLoRA, and DoRA through specific CLI commands. This skill provides the exact command-line instructions needed to start SFT or CPT workflows on your target model and dataset.
Yes, ms-swift supports GRPO-family RLHF training. This skill provides the necessary CLI commands and reference docs to configure and run reinforcement learning workflows for your models.
Yes, ms-swift includes Megatron distributed training context for performance optimization. This skill guides you through the distributed workflows covered by the framework.
ms-swift allows deployment as an OpenAI-compatible API. This skill provides the inference and deployment commands needed to serve your model using engines like vLLM, SGLang, or LmDeploy.
ms-swift supports AWQ, GPTQ, FP8, and BNB quantization options. This skill aligns your model export and optimization steps with these covered capabilities to reduce model size.
ms-swift usage is limited to its documented CLI commands and covered capabilities. You must check the referenced docs under references/ before answering or executing training, inference, and export steps.
Aniket-a14Train language models with TRL on Hugging Face Jobs.
vasilyu1983Optimize LLM inference performance and cost with vLLM and quantization.
pacphiDefine and train PyTorch neural networks with CUDA acceleration.
DaddyElonMusk69Fine-tunes LLMs with TRL using SFT, DPO, and PPO/GRPO pipelines.
