llm-jp-3-13b-finetune

実行手順

<!-- Provide a quick summary of what the model is/does. -->

以下はHugging Face上のモデルを用いてELYZA-tasks-100-TVの出力を得るためのコードです。

使用方法

１．必要なライブラリのインストールを行います。

<!-- Provide the basic links for the model. -->

!pip install -U bitsandbytes
!pip install -U transformers
!pip install -U accelerate
!pip install -U datasets
!pip install -U peft

２．Hugging Faceのトークンを取得していることを確認してください。以下はGoogle Colabでuserdataをインポートします。

<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->

from google.colab import userdata
HF_TOKEN = userdata.get('HF_TOKEN')

モデル・トークナイザの読み込み

from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    BitsAndBytesConfig,
)
from peft import PeftModel
import torch
from tqdm import tqdm
import json
import re

model_id = "llm-jp/llm-jp-3-13b"
adapter_id = "kenskit/llm-jp-3-13b-finetune"

# QLoRA用の設定
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16,
)

# モデル読み込み
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    quantization_config=bnb_config,
    device_map="auto",
    token=HF_TOKEN
)

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True, token=HF_TOKEN)

# Peftモデルを適用
model = PeftModel.from_pretrained(model, adapter_id, token=HF_TOKEN)

入力データ

datasets = []
with open("./elyza-tasks-100-TV_0.jsonl", "r") as f:
    item = ""
    for line in f:
      line = line.strip()
      item += line
      if item.endswith("}"):
        datasets.append(json.loads(item))
        item = ""

推論実行

<!-- Relevant interpretability work for the model goes here -->

results = []
for data in tqdm(datasets):
    input_data = data["input"]

    prompt = f"""### 指示
{input_data}
### 回答
"""

    tokenized_input = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt").to(model.device)
    attention_mask = torch.ones_like(tokenized_input)
    with torch.no_grad():
        outputs = model.generate(
            tokenized_input,
            attention_mask=attention_mask,
            max_new_tokens=100,
            do_sample=False,
            repetition_penalty=1.2,
            pad_token_id=tokenizer.eos_token_id
        )[0]

    output = tokenizer.decode(outputs[tokenized_input.size(1):], skip_special_tokens=True)
    
    results.append({
        "input": input_data,
        "output": output
    })

出力の保存

jsonl_id = re.sub(".*/", "", adapter_id)
with open(f"./{jsonl_id}-outputs.jsonl", 'w', encoding='utf-8') as f:
    for result in results:
        json.dump(result, f, ensure_ascii=False)
        f.write('\n')