【2026年最新】LLMファインチューニング実践ガイド：LoRA・QLoRA・フルファインチューニングの使い分け

pip install transformers peft trl datasets accelerate bitsandbytes

from datasets import Dataset

# 学習データの形式
train_data = [
    {
        "instruction": "以下の文章を要約してください",
        "input": "量子コンピューティングは...(長文)",
        "output": "量子コンピューティングは...（要約）"
    },
    # ... 1000件以上推奨
]

dataset = Dataset.from_list(train_data)
dataset = dataset.train_test_split(test_size=0.1)

from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from peft import LoraConfig, get_peft_model, prepare_model_for_kbit_training
from trl import SFTTrainer, SFTConfig

# モデル読み込み（QLoRA用4bit量子化）
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype="bfloat16",
    bnb_4bit_use_double_quant=True,
)

model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    quantization_config=bnb_config,
    device_map="auto",
)
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.1-8B-Instruct")
tokenizer.pad_token = tokenizer.eos_token

# LoRA設定
lora_config = LoraConfig(
    r=16,
    lora_alpha=32,
    target_modules=[
        "q_proj", "k_proj", "v_proj", "o_proj",
        "gate_proj", "up_proj", "down_proj",
    ],
    lora_dropout=0.05,
    bias="none",
    task_type="CAUSAL_LM",
)

model = prepare_model_for_kbit_training(model)
model = get_peft_model(model, lora_config)

model.print_trainable_parameters()
# trainable params: 13,631,488 || all params: 8,043,901,440 || trainable%: 0.17

# トレーニング
training_config = SFTConfig(
    output_dir="./output",
    num_train_epochs=3,
    per_device_train_batch_size=4,
    gradient_accumulation_steps=4,
    learning_rate=2e-4,
    lr_scheduler_type="cosine",
    warmup_ratio=0.1,
    logging_steps=10,
    save_strategy="epoch",
    bf16=True,
    max_seq_length=2048,
)

trainer = SFTTrainer(
    model=model,
    train_dataset=dataset["train"],
    eval_dataset=dataset["test"],
    args=training_config,
    tokenizer=tokenizer,
)

trainer.train()
trainer.save_model("./my-lora-adapter")

from peft import PeftModel

base_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B-Instruct",
    device_map="auto",
)

model = PeftModel.from_pretrained(base_model, "./my-lora-adapter")

# マージして単体モデルとして保存
merged_model = model.merge_and_unload()
merged_model.save_pretrained("./merged-model")