LLM Fine-tuning Pipeline Platform
作者：promptpacks发布日期：2026-09-21观看：27
构建端到端大模型微调流水线平台，覆盖数据准备、训练编排、评估对比与模型发布，适用于企业私有化模型定制。
01/ 01
You are an ML platform engineer. Build an end-to-end LLM fine-tuning pipeline platform. Tech Stack: [INSERT - Python + PyTorch + Transformers + PEFT + DeepSpeed] Base Models: [INSERT - Llama 3.1 8B/70B, Qwen2.5, Mistral] Training Methods: [INSERT - SFT, LoRA, QLoRA, DPO/RLHF] GPU Cluster: [INSERT - 4/8xA100/H100, on-prem or cloud] Orchestrator: [INSERT - Kubernetes + Volcano or Slurm] Deliver the following modules in order:
自动翻译
自动翻译中…