Which fine-tuning methods are supported?

The skill supports various alignment and fine-tuning techniques including SFT, ORPO, DPO, LoRA, QLoRA, and DoRA.

What is the LLM Fine-tuning Coach?

It is a specialized Claude Code skill that provides a guided, coach-style workflow for training, evaluating, and deploying fine-tuned AI models.

How does the coach-like guidance work?

It uses specialized agents to proactively clarify project goals, diagnose performance issues, and recommend data sources rather than making assumptions.

Can I track different versions of my models?

Yes, it includes a robust versioning system that tracks data lineage, configuration snapshots, and evaluation results for every iteration.

Does it support remote training environments?

Yes, it supports hybrid environments including local GPU setups, remote SSH servers, and major cloud providers like AWS or GCP.

LLM Fine-tuning Coach

Name: LLM Fine-tuning Coach
Author: p988744

byp988744

数据科学与机器学习

Guides users through the end-to-end Large Language Model fine-tuning lifecycle using a coach-driven workflow.

关于

The LLM Fine-tuning Coach is a comprehensive workflow engine for Claude Code designed to simplify complex model optimization tasks. It employs a coach-like approach to clarify project goals, manage multiple training tasks, and maintain strict data lineage. From data pipeline configuration and automated benchmark evaluation to deploying fine-tuned models on HuggingFace, it provides a structured framework supporting advanced techniques like LoRA, QLoRA, and DPO, ensuring every training iteration is reproducible and well-documented.

主要功能

Automated data pipeline configuration with multi-source support
Full version lineage tracking for data, configs, and model performance
Coach-driven goal clarification and project scoping agents
Seamless HuggingFace deployment and benchmark evaluation integration
End-to-end management of LoRA, QLoRA, and DoRA training workflows
0 GitHub stars

使用场景

Establishing reproducible data pipelines for continuous model improvement
Comparing multiple model versions and hyperparameter configurations for optimization
Training domain-specific models with proprietary or synthesized datasets

关于

主要功能

Automated data pipeline configuration with multi-source support
Full version lineage tracking for data, configs, and model performance
Coach-driven goal clarification and project scoping agents
Seamless HuggingFace deployment and benchmark evaluation integration
End-to-end management of LoRA, QLoRA, and DoRA training workflows
0 GitHub stars

使用场景

Establishing reproducible data pipelines for continuous model improvement
Comparing multiple model versions and hyperparameter configurations for optimization
Training domain-specific models with proprietary or synthesized datasets