llama3.1 파인튜닝 퀵하게~ (with Colab)

Unsloth 라는 기가막힌 라이브러리가 있다고 하여 바로 실행 해봤다. 

 

아래 깃허브에서 원하는 모델 선택해서 Colab에서 바로 실행해 볼 수도 있음~!!

 

https://github.com/unslothai/unsloth?tab=readme-ov-file

GitHub - unslothai/unsloth: Finetune Llama 3.1, Mistral, Phi & Gemma LLMs 2-5x faster with 80% less memory

 

실제 실행해본 코드 첨부 해봅니다. 

 

## 필요 라이브러리 설치

%%capture
# Installs Unsloth, Xformers (Flash Attention) and all other packages!
!pip install "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
!pip install --no-deps "xformers==0.0.27" trl peft accelerate bitsandbytes


## Unsloth 셋팅
from unsloth import FastLanguageModel
import torch
max_seq_length = 2048 # Choose any! We auto support RoPE Scaling internally!
dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.

# 4bit pre quantized models we support for 4x faster downloading + no OOMs.
fourbit_models = [
    "unsloth/Meta-Llama-3.1-8B-bnb-4bit",      # Llama-3.1 15 trillion tokens model 2x faster!
    "unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit",
    "unsloth/Meta-Llama-3.1-70B-bnb-4bit",
    "unsloth/Meta-Llama-3.1-405B-bnb-4bit",    # We also uploaded 4bit for 405b!
    "unsloth/Mistral-Nemo-Base-2407-bnb-4bit", # New Mistral 12b 2x faster!
    "unsloth/Mistral-Nemo-Instruct-2407-bnb-4bit",
    "unsloth/mistral-7b-v0.3-bnb-4bit",        # Mistral v3 2x faster!
    "unsloth/mistral-7b-instruct-v0.3-bnb-4bit",
    "unsloth/Phi-3.5-mini-instruct",           # Phi-3.5 2x faster!
    "unsloth/Phi-3-medium-4k-instruct",
    "unsloth/gemma-2-9b-bnb-4bit",
    "unsloth/gemma-2-27b-bnb-4bit",            # Gemma 2x faster!
] # More models at https://huggingface.co/unsloth

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,
    # token = "hf_...", # use one if using gated models like meta-llama/Llama-2-7b-hf
)

model = FastLanguageModel.get_peft_model(
    model,
    r = 16, # Choose any number > 0 ! Suggested 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, # Supports any, but = 0 is optimized
    bias = "none",    # Supports any, but = "none" is optimized
    # [NEW] "unsloth" uses 30% less VRAM, fits 2x larger batch sizes!
    use_gradient_checkpointing = "unsloth", # True or "unsloth" for very long context
    random_state = 3407,
    use_rslora = False,  # We support rank stabilized LoRA
    loftq_config = None, # And LoftQ
)

## Data 셋팅 (Teddy 님 강의에서 사용한 Data 활용)

alpaca_prompt = """Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
{}

### Response:
{}"""

EOS_TOKEN = tokenizer.eos_token # Must add EOS_TOKEN
def formatting_prompts_func(examples):
    instructions = examples["instruction"]
    outputs      = examples["output"]
    texts = []
    for instruction, output in zip(instructions, outputs):
        # Must add EOS_TOKEN, otherwise your generation will go on forever!
        text = alpaca_prompt.format(instruction, output) + EOS_TOKEN
        texts.append(text)
    return { "text" : texts, }
pass

from datasets import load_dataset
dataset = load_dataset("teddylee777/QA-Dataset-mini", split = "train")
dataset = dataset.map(formatting_prompts_func, batched = True,)

## 학습 준비
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, # Can make training 5x faster for short sequences.
    args = TrainingArguments(
        per_device_train_batch_size = 2,
        gradient_accumulation_steps = 4,
        warmup_steps = 5,
        # num_train_epochs = 1, # Set this for 1 full training run.
        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",
    ),
)

## 학습
trainer_stats = trainer.train() 

# 추론해봄
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
inputs = tokenizer(
[
    alpaca_prompt.format(
        "CES 2024의 주제에 대해서 말해줘.", # instruction
        "", # output - leave this blank for generation!
    )
], return_tensors = "pt").to("cuda")

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

 

학습에 사용 했던 데이터

 

학습 후 추론해본 데이터

 

그리고 그냥 llama 3.1 기본 모델에 똑같은 질문을 해봄

 

 

확실히 파인튜닝이 된 것으로 보임. 

 

이제 내 데이터 잘 만들어서 해봐야겠다~ 

문서에서 llama 학습용 Data 만드는 법과 파인튜닝은 아래 영상 참고했습니다~

감사합니다.

 

https://www.youtube.com/watch?v=oZY0D8N6bC8