fine-tuning-expert
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 许可
- MIT
- 最近更新
- 2026-09-01
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- fine-tuning-expert
- license
- MIT
- metadata
- { "role": "expert", "scope": "implementation", "author": "https://github.com/Jeffallan", "domain": "data-ml", "version": "1.1.0", "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", "output-format": "code", "related-skills": "devops-engineer" }
- 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.
02 资源文件 5
- 参考文档references/dataset-preparation.md16.1 KB
- 参考文档references/deployment-optimization.md16.7 KB
- 参考文档references/evaluation-metrics.md17.6 KB
- 参考文档references/hyperparameter-tuning.md16.2 KB
- 参考文档references/lora-peft.md10.8 KB
03 SKILL.md 正文
Fine-Tuning Expert
Senior ML engineer specializing in LLM fine-tuning, parameter-efficient methods, and production model optimization.
Core Workflow
- 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
- Checkpoint:
- 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
- Training — Configure hyperparameters, monitor loss curves, checkpoint regularly
- Checkpoint: validation loss must decrease; plateau or increase signals overfitting
- 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
- 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
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:
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:
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:
- Dataset preparation script with validation logic (schema checks, token-length histogram, deduplication)
- Training configuration (full
TrainingArguments+LoraConfigblock, commented) - Evaluation script reporting perplexity, task-specific metrics, and latency
- Brief design rationale — why this PEFT method, rank, and learning rate were chosen for this task
04 同仓库更多 Skills
wordpress-pro
Develops custom WordPress themes and plugins, creates and registers Gutenberg blocks and block patterns, configures WooCommerce stores, implements WordPress REST API endpoints, applies security hardening (nonces, sanitization, escaping, capability checks), and optimizes performance through caching and query tuning. Use when building WordPress themes, writing plugins, customizing Gutenberg blocks, extending WooCommerce, working with ACF, using the WordPress REST API, applying hooks and filters, or improving WordPress performance and security.
websocket-engineer
Use when building real-time communication systems with WebSockets or Socket.IO. Invoke for bidirectional messaging, horizontal scaling with Redis, presence tracking, room management.
webapp-testing
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
web-design-guidelines
UI/UX best practices and accessibility audit. Use when reviewing UI code, checking accessibility, running accessibility audits, auditing forms, or ensuring web interface best practices. Triggers on "audit accessibility", "check WCAG", "review UI", "check accessibility", "audit design", "review UX", or "check best practices".
web-artifacts-builder
Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn/ui). Use for complex artifacts requiring state management, routing, or shadcn/ui components - not for simple single-file HTML/JSX artifacts.
vue-expert
Builds Vue 3 components with Composition API patterns, configures Nuxt 3 SSR/SSG projects, sets up Pinia stores, scaffolds Quasar/Capacitor mobile apps, implements PWA features, and optimises Vite builds. Use when creating Vue 3 applications with Composition API, writing reusable composables, managing state with Pinia, building hybrid mobile apps with Quasar or Capacitor, configuring service workers, or tuning Vite configuration and TypeScript integration.