簡単にLlama3.1をFine Tuningする

必要なパッケージをインストールします

%%capture
!pip install "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
!pip install --no-deps "xformers<0.0.27" "trl<0.9.0" peft accelerate bitsandbytes

Unsloth初期化

from unsloth import FastLanguageModel
import torch
max_seq_length = 2048
dtype = None # 自動検出ではNoneです。 Float16はTesla T4、V100、Bfloat16はAmpere+用です
load_in_4bit = True # メモリ使用量を減らすには、4 ビット量子化を使用します。 False も可能。

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "unsloth/Meta-Llama-3.1-8B",
    max_seq_length = max_seq_length,
    dtype = dtype,
    load_in_4bit = load_in_4bit,
)

layers patch

model = FastLanguageModel.get_peft_model(
    model,
    r = 16, # 任意の数字 > 0 を選択します。推奨 8、16、32、64、128
    target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
                      "gate_proj", "up_proj", "down_proj",],
    lora_alpha = 16,
    lora_dropout = 0, 
    bias = "none",
    use_gradient_checkpointing = "unsloth", # 「true」または「unsloth」は、非常に長い文脈のためです
    random_state = 3407,
    use_rslora = False, 
    loftq_config = None,
)

データセット取ってくる

alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{}

### Input:
{}

### Response:
{}"""

EOS_TOKEN = tokenizer.eos_token # EOS_TOKEN を追加する必要があります
def formatting_prompts_func(examples):
    instructions = examples["instruction"]
    inputs       = examples["input"]
    outputs      = examples["output"]
    texts = []
    for instruction, input, output in zip(instructions, inputs, outputs):
        text = alpaca_prompt.format(instruction, input, output) + EOS_TOKEN
        texts.append(text)
    return { "text" : texts, }
pass

from datasets import load_dataset
dataset = load_dataset("yahma/alpaca-cleaned", split = "train") # 他のデータセットも可能です
dataset = dataset.map(formatting_prompts_func, batched = True,)

SFTTrainer

from trl import SFTTrainer
from transformers import TrainingArguments
from unsloth import is_bfloat16_supported

trainer = SFTTrainer(
    model = model,
    tokenizer = tokenizer,
    train_dataset = dataset,
    dataset_text_field = "text",
    max_seq_length = max_seq_length,
    dataset_num_proc = 2,
    packing = False,
    args = TrainingArguments(
        per_device_train_batch_size = 2,
        gradient_accumulation_steps = 4,
        warmup_steps = 5,
        # num_train_epochs = 1、# これを 1をすると、完全なトレーニング実行に設定します。
        max_steps = 60,
        learning_rate = 2e-4,
        fp16 = not is_bfloat16_supported(),
        bf16 = is_bfloat16_supported(),
        logging_steps = 1,
        optim = "adamw_8bit",
        weight_decay = 0.01,
        lr_scheduler_type = "linear",
        seed = 3407,
        output_dir = "outputs",
    ),
)

現在のメモリの統計を表示

#@title 現在のメモリの統計を表示
gpu_stats = torch.cuda.get_device_properties(0)
start_gpu_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
max_memory = round(gpu_stats.total_memory / 1024 / 1024 / 1024, 3)
print(f"GPU = {gpu_stats.name}. Max memory = {max_memory} GB.")
print(f"{start_gpu_memory} GB of memory reserved.")

Train！！

trainer_stats = trainer.train()

最終的なメモリと時刻の統計情報を表示

#@title 最終的なメモリと時刻の統計情報を表示
used_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
used_memory_for_lora = round(used_memory - start_gpu_memory, 3)
used_percentage = round(used_memory         /max_memory*100, 3)
lora_percentage = round(used_memory_for_lora/max_memory*100, 3)
print(f"{trainer_stats.metrics['train_runtime']} seconds used for training.")
print(f"{round(trainer_stats.metrics['train_runtime']/60, 2)} minutes used for training.")
print(f"Peak reserved memory = {used_memory} GB.")
print(f"Peak reserved memory for training = {used_memory_for_lora} GB.")
print(f"Peak reserved memory % of max memory = {used_percentage} %.")
print(f"Peak reserved memory for training % of max memory = {lora_percentage} %.")

model作る

FastLanguageModel.for_inference(model) # Enable native 2x faster inference
inputs = tokenizer(
[
    alpaca_prompt.format(
        "Continue the fibonnaci sequence.", # 命令
        "1, 1, 2, 3, 5, 8", # input
        "", # Output - 生成のために空白のままにしておいてください！
    )
], return_tensors = "pt").to("cuda")

outputs = model.generate(**inputs, max_new_tokens = 64, use_cache = True)
tokenizer.batch_decode(outputs)

FastLanguageModel.for_inference(model) 
inputs = tokenizer(
[
    alpaca_prompt.format(
        "Continue the fibonnaci sequence.", # 命令
        "1, 1, 2, 3, 5, 8", # input
        "", # Output - 生成のために空白のままにしておいてください！
    )
], return_tensors = "pt").to("cuda")

from transformers import TextStreamer
text_streamer = TextStreamer(tokenizer)
_ = model.generate(**inputs, streamer = text_streamer, max_new_tokens = 128)

ローカル保存

model.save_pretrained("lora_model")
tokenizer.save_pretrained("lora_model")

オンライン保存

model.push_to_hub("UserName/lora_model", token = "YOUR_TOKEN") 
tokenizer.push_to_hub("UserName/lora_model", token = "YOUR_TOKEN") 

Colabリンク

Colab コード