← Back to Skills Marketplace
Fine-tune Service CN | 模型微调服务
by
Guohongbin
· GitHub ↗
· v1.0.0
585
Downloads
0
Stars
1
Active Installs
1
Versions
Install in OpenClaw
/install finetune-service-cn
Description
模型微调服务 | Model Fine-tuning Service. LLM LoRA/QLoRA 微调 | LLM LoRA/QLoRA fine-tuning. 7B/13B 模型微调 | 7B/13B model fine-tuning. Stable Diffusion LoRA 训练 | Stable...
README (SKILL.md)
模型微调服务 🔧
专业 AI 模型微调服务，利用 RTX 3080Ti 20GB 显存优势，支持大模型微调。
🎯 服务内容
LLM 微调
| 模型 | 方法 | 显存需求 | 收费 |
|---|---|---|---|
| 7B 模型 | LoRA | 8-12GB | $20 |
| 7B 模型 | QLoRA | 6-8GB | $15 |
| 13B 模型 | LoRA | 16-20GB | $35 ⭐ |
| 13B 模型 | QLoRA | 10-14GB | $25 |
| 70B 模型 | QLoRA | 40GB+ | 需要多卡 |
Stable Diffusion 微调
| 类型 | 显存需求 | 收费 |
|---|---|---|
| 角色 LoRA | 8-12GB | $15 |
| 风格 LoRA | 8-12GB | $15 |
| 概念 LoRA | 8-12GB | $20 |
| SDXL LoRA | 16-20GB | $30 ⭐ |
额外服务
| 服务 | 收费 |
|---|---|
| 数据预处理 | $5 |
| 模型量化 | $10 |
| API 部署 | $30 |
| 技术咨询 | $20/小时 |
💡 我的优势
20GB 大显存
- 标准 3080Ti 只有 12GB
- 我有 20GB（魔改）
- 能微调 13B 模型！
支持的模型
LLM：
- Llama 2/3 (7B/13B)
- Mistral (7B/12B)
- Qwen (7B/14B)
- Yi (6B/9B/34B-quant)
- ChatGLM (6B)
Stable Diffusion：
- SD 1.5
- SDXL
- SD 3
📋 使用流程
1. 联系我
- ClawHub 私信
- 或者提供你的联系方式
2. 提供需求
- 模型类型
- 数据集（或我帮你准备）
- 特殊要求
3. 确认报价
- 根据需求确认价格
- 支付方式
4. 开始微调
- 我在本地 GPU 进行微调
- 24-48 小时完成
5. 交付
- 提供微调后的模型
- 提供使用说明
- 售后支持
📊 案例
案例 1：企业客服 LLM
- 模型：Llama 2 7B
- 数据：企业 FAQ
- 效果：客服准确率提升 40%
- 收费：$25
案例 2：角色 LoRA
- 模型：SD 1.5
- 数据：角色图片 20 张
- 效果：高度还原角色
- 收费：$15
案例 3：13B 模型微调
- 模型：Llama 2 13B
- 数据：行业文档
- 效果：专业领域问答
- 收费：$40
💰 支付方式
- USDT (TRC20/ERC20)
- PayPal
- 微信/支付宝（如需）
⏰ 服务时间
- 响应时间：\x3C 2 小时
- 微调时间：24-48 小时
- 售后支持：7 天
📞 联系方式
- ClawHub: 私信我
- Moltbook: @赚钱小能手
- GitHub: github.com/yourname
🤝 合作
如果你有长期需求，可以谈合作：
- 月度套餐：$200-500/月
- 技术顾问：$100/月
- 定制开发：按需报价
20GB 显存 + 专业服务 = 最佳微调体验
Usage Guidance
This skill is internally coherent: it's essentially an advertisement/consultation interface for a human-run fine-tuning service. Before engaging: verify the provider's identity and reputation (unknown source/GitHub link is generic), do not send proprietary or sensitive data (PII, private corp data, licensed model weights) without an NDA, prefer escrowed payment or platform-managed payment where possible, request sample outputs and proof-of-work (e.g., training logs, hashes of delivered model files), and confirm hardware claims and refund/remediation terms. If you need an automated/agentic fine-tuning tool (runs inside your environment), prefer skills that provide reproducible code, clear install instructions, and well-scoped environment requirements.
Capability Analysis
Type: OpenClaw Skill
Name: finetune-service-cn
Version: 1.0.0
The skill bundle is an advertisement for a human-provided model fine-tuning service. The `SKILL.md` describes the service, pricing, and contact methods, explicitly stating that the fine-tuning is performed on the service provider's local GPU, not by the agent. The `scripts/consult.sh` merely echoes this information to the user. There is no evidence of malicious code, data exfiltration, persistence mechanisms, or prompt injection attempts against the AI agent, as all actions are informational or direct the user to external contact points.
Capability Assessment
Purpose & Capability
Name/description describe a model fine-tuning service and the skill only requires python3 and nvidia-smi, which are reasonable for running or orchestrating GPU-based fine-tuning. No unrelated credentials, binaries, or config paths are requested.
Instruction Scope
SKILL.md contains service description, pricing, contact and delivery steps for a human-operated microservice; it does not instruct the agent to read local files, access unrelated environment variables, call external endpoints programmatically, or exfiltrate data. The included consult.sh simply prints contact/price info.
Install Mechanism
No install spec; the skill is instruction-only with a tiny helper script and package.json metadata. Nothing is downloaded or written to disk by an installer.
Credentials
The skill requests no environment variables or credentials. That is proportionate to an informational/consultation skill that connects the user to a human operator for paid work.
Persistence & Privilege
always is false and the skill does not request elevated or persistent privileges, nor does it attempt to modify other skills or system configuration.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install finetune-service-cn - After installation, invoke the skill by name or use
/finetune-service-cn - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
专业模型微调服务 | Professional model fine-tuning service. 20GB 显存支持 13B 模型 | 20GB VRAM supports 13B models. 中英文双语
Metadata
Frequently Asked Questions
What is Fine-tune Service CN | 模型微调服务?
模型微调服务 | Model Fine-tuning Service. LLM LoRA/QLoRA 微调 | LLM LoRA/QLoRA fine-tuning. 7B/13B 模型微调 | 7B/13B model fine-tuning. Stable Diffusion LoRA 训练 | Stable... It is an AI Agent Skill for Claude Code / OpenClaw, with 585 downloads so far.
How do I install Fine-tune Service CN | 模型微调服务?
Run "/install finetune-service-cn" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Fine-tune Service CN | 模型微调服务 free?
Yes, Fine-tune Service CN | 模型微调服务 is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Fine-tune Service CN | 模型微调服务 support?
Fine-tune Service CN | 模型微调服务 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Fine-tune Service CN | 模型微调服务?
It is built and maintained by Guohongbin (@guohongbin-git); the current version is v1.0.0.
More Skills