研究与检索SKILL.md
模型微调方案研究
v1比较训练数据、微调方法和评估过程,明确实验的成本与约束。
技能概述
这项技能做什么
比较训练数据、微调方法和评估过程,明确实验的成本与约束。
对应原始技能标识:fine-tuning-expert。本站整理中文说明,技能原文由 Jeffallan/claude-skills 维护。
适用场景与输入
分类:研究与检索。建议准备:研究问题、已知资料、比较对象、证据范围和判断标准。
获取与使用
- 打开固定版本的完整技能目录,按仓库说明获取技能。
- 原目录包含
references等配套资源。请获取完整技能目录,单独下载本站的 SKILL.md 不包含这些文件。 - 在 Claude Code 或具备相应工具能力的兼容 Agent 中加载;兼容 SKILL.md 格式不代表各平台工具完全一致。
- 参照页面的使用示例提供材料,按照原文步骤执行,并核对实际输出。
输出与验证
区分事实、假设与推断;按证据组织结论,列出局限和下一步验证建议。本站未执行来源仓库中的脚本,也不把示例当作已经完成的运行结果。
来源与许可
- 原始 SKILL.md
- 完整技能目录及配套资源
- 固定提交:
078a25cc5bfd8011459be7b91d004e900c1d02d3 - 许可证:MIT;技能原文和下载内容中保留以下版权与许可说明。
MIT License
Copyright (c) 2025
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
如何使用这个 Skill
下载 SKILL.md,查看文件中的说明与依赖,再按你的 Agent 支持的方式导入。
SKILL.mdMarkdown · 1 个文件
SKILL.md
---
name: fine-tuning-expert
description: "Use when fine-tuning LLMs, training custom models, or adapting foundation models for specific tasks. Invoke for configuring LoRA/QLoRA adapters, preparing JSONL training datasets, setting hyperparameters for fine-tuning runs, adapter training, transfer learning, finetuning with Hugging Face PEFT, OpenAI fine-tuning, instruction tuning, RLHF, DPO, or quantizing and deploying fine-tuned models. Trigger terms include: LoRA, QLoRA, PEFT, finetuning, fine-tuning, adapter tuning, LLM training, model training, custom model."
license: MIT
metadata:
author: https://github.com/Jeffallan
company: https://synergetic.solutions
version: "1.1.0"
domain: data-ml
triggers: fine-tuning, fine tuning, finetuning, LoRA, QLoRA, PEFT, adapter tuning, transfer learning, model training, custom model, LLM training, instruction tuning, RLHF, model optimization, quantization
role: expert
scope: implementation
output-format: code
related-skills: devops-engineer
---
# Fine-Tuning Expert
Senior ML engineer specializing in LLM fine-tuning, parameter-efficient methods, and production model optimization.
## Core Workflow
1. **Dataset preparation** — Validate and format data; run quality checks before training starts
- Checkpoint: `python validate_dataset.py --input data.jsonl` — fix all errors before proceeding
2. **Method selection** — Choose PEFT technique based on GPU memory and task requirements
- Use LoRA for most tasks; QLoRA (4-bit) when GPU memory is constrained; full fine-tune only for small models
3. **Training** — Configure hyperparameters, monitor loss curves, checkpoint regularly
- Checkpoint: validation loss must decrease; plateau or increase signals overfitting
4. **Evaluation** — Benchmark against the base model; test on held-out set and edge cases
- Checkpoint: collect perplexity, task-specific metrics (BLEU/ROUGE), and latency numbers
5. **Deployment** — Merge adapter weights, quantize, measure inference throughput before serving
## Reference Guide
Load detailed guidance based on context:
| Topic | Reference | Load When |
|-------|-----------|-----------|
| LoRA/PEFT | `references/lora-peft.md` | Parameter-efficient fine-tuning, adapters |
| Dataset Prep | `references/dataset-preparation.md` | Training data formatting, quality checks |
| Hyperparameters | `references/hyperparameter-tuning.md` | Learning rates, batch sizes, schedulers |
| Evaluation | `references/evaluation-metrics.md` | Benchmarking, metrics, model comparison |
| Deployment | `references/deployment-optimization.md` | Model merging, quantization, serving |
## Minimal Working Example — LoRA Fine-Tuning with Hugging Face PEFT
```python
from datasets import load_dataset
from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments
from peft import LoraConfig, get_peft_model, TaskType
from trl import SFTTrainer
import torch
# 1. Load base model and tokenizer
model_id = "meta-llama/Llama-3-8B"
tokenizer = AutoTokenizer.from_pretrained(model_id)
tokenizer.pad_token = tokenizer.eos_token
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype=torch.bfloat16,
device_map="auto",
)
# 2. Configure LoRA adapter
lora_config = LoraConfig(
task_type=TaskType.CAUSAL_LM,
r=16, # rank — increase for more capacity, decrease to save memory
lora_alpha=32, # scaling factor; typically 2× rank
target_modules=["q_proj", "v_proj"],
lora_dropout=0.05,
bias="none",
)
model = get_peft_model(model, lora_config)
model.print_trainable_parameters() # verify: should be ~0.1–1% of total params
# 3. Load and format dataset (Alpaca-style JSONL)
dataset = load_dataset("json", data_files={"train": "train.jsonl", "test": "test.jsonl"})
def format_prompt(example):
return {"text": f"### Instruction:\n{example['instruction']}\n\n### Response:\n{example['output']}"}
dataset = dataset.map(format_prompt)
# 4. Training arguments
training_args = TrainingArguments(
output_dir="./checkpoints",
num_train_epochs=3,
per_device_train_batch_size=4,
gradient_accumulation_steps=4, # effective batch size = 16
learning_rate=2e-4,
lr_scheduler_type="cosine",
warmup_ratio=0.03, # always use warmup
fp16=False,
bf16=True,
logging_steps=10,
eval_strategy="steps",
eval_steps=100,
save_steps=200,
load_best_model_at_end=True,
)
# 5. Train
trainer = SFTTrainer(
model=model,
args=training_args,
train_dataset=dataset["train"],
eval_dataset=dataset["test"],
dataset_text_field="text",
max_seq_length=2048,
)
trainer.train()
# 6. Save adapter weights only
model.save_pretrained("./lora-adapter")
tokenizer.save_pretrained("./lora-adapter")
```
**QLoRA variant** — add these lines before loading the model to enable 4-bit quantization:
```python
from transformers import BitsAndBytesConfig
bnb_config = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_quant_type="nf4",
bnb_4bit_compute_dtype=torch.bfloat16,
bnb_4bit_use_double_quant=True,
)
model = AutoModelForCausalLM.from_pretrained(model_id, quantization_config=bnb_config, device_map="auto")
```
**Merge adapter into base model for deployment:**
```python
from peft import PeftModel
base = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16)
merged = PeftModel.from_pretrained(base, "./lora-adapter").merge_and_unload()
merged.save_pretrained("./merged-model")
```
## Constraints
### MUST DO
- Validate dataset quality before training
- Use parameter-efficient methods for large models (>7B)
- Monitor training/validation loss curves
- Document hyperparameters and training config
- Version datasets and model checkpoints
- Always include a learning rate warmup
### MUST NOT DO
- Skip data quality validation
- Overfit on small datasets — use regularisation (dropout, weight decay) and early stopping
- Merge incompatible adapters (mismatched rank, base model, or target modules)
- Deploy without evaluation against a held-out set and latency benchmark
## Output Templates
When implementing fine-tuning, always provide:
1. **Dataset preparation script** with validation logic (schema checks, token-length histogram, deduplication)
2. **Training configuration** (full `TrainingArguments` + `LoraConfig` block, commented)
3. **Evaluation script** reporting perplexity, task-specific metrics, and latency
4. **Brief design rationale** — why this PEFT method, rank, and learning rate were chosen for this task
Maintained by [@jeffallan](https://github.com/jeffallan), Principal Consultant at [Synergetic Solutions](https://synergetic.solutions)
[Documentation](https://jeffallan.github.io/claude-skills/skills/data-ml/fine-tuning-expert/)
---
## Source and license
Original skill: https://github.com/Jeffallan/claude-skills/blob/078a25cc5bfd8011459be7b91d004e900c1d02d3/skills/fine-tuning-expert/SKILL.md
Complete source directory: https://github.com/Jeffallan/claude-skills/tree/078a25cc5bfd8011459be7b91d004e900c1d02d3/skills/fine-tuning-expert
This catalog copy preserves the original instructions. Supporting scripts and resources are available in the original directory; this single-file export does not bundle them.
MIT License
Copyright (c) 2025
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
使用示例
请使用「模型微调方案研究」技能处理下面的任务。 目标:比较训练数据、微调方法和评估过程,明确实验的成本与约束。 输入材料:研究问题、已知资料、比较对象、证据范围和判断标准。 输出要求:区分事实、假设与推断;按证据组织结论,列出局限和下一步验证建议。 请先核对所需资料和工具;缺少关键信息时提出具体问题。涉及对外发布、删除或生产环境变更时,先提供可审查的计划,获得授权后再执行。

关注官方公众号
抢先获取热门 Skill 实操干货、产品更新与官方活动资讯。

加入官方交流群
对接创作者与同好,交流技能开发,共建 AI 技能生态。