Update README.md

README.md

@@ -20,3 +20,114 @@ language:
 This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
 
 [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
+
+# 提出したjsonlファイルの出力方法
+
+1. 必要なライブラリのインストール
+  ```
+  pip install unsloth
+  pip uninstall unsloth -y && pip install --upgrade --no-cache-dir "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
+  ```
+
+1. 下記のコードを実行
+  ```
+  import json
+  from dataclasses import dataclass
+  from pathlib import Path
+  from typing import Any, Dict, List
+  
+  import torch
+  from tqdm import tqdm
+  from unsloth import FastLanguageModel
+  
+  
+  @dataclass
+  class ModelConfig:
+      model_name: str = "imagfff/llm-jp-3-13b-it"
+      max_seq_length: int = 2048
+      dtype: Any = None
+      load_in_4bit: bool = True
+      token: str = "HF token"
+  
+  
+  def load_model(config: ModelConfig) -> tuple[Any, Any]:
+      """モデルとトークナイザーを読み込む"""
+      model, tokenizer = FastLanguageModel.from_pretrained(
+          model_name=config.model_name,
+          max_seq_length=config.max_seq_length,
+          dtype=config.dtype,
+          load_in_4bit=config.load_in_4bit,
+          token=config.token,
+      )
+      FastLanguageModel.for_inference(model)
+      return model, tokenizer
+  
+  
+  def load_datasets(file_path: str) -> List[Dict[str, Any]]:
+      """JSONLファイルからデータセットを読み込む"""
+      datasets = []
+      try:
+          with open(file_path) as f:
+              item = ""
+              for line in f:
+                  line = line.strip()
+                  item += line
+                  if item.endswith("}"):
+                      datasets.append(json.loads(item))
+                      item = ""
+          return datasets
+      except (FileNotFoundError, json.JSONDecodeError) as e:
+          raise Exception(f"データセットの読み込みに失敗しました: {e}") from e
+  
+  
+  def generate_prediction(model: Any, tokenizer: Any, input_text: str) -> str:
+      """モデルによる推論を実行"""
+      prompt = f"### 指示\n{input_text}\n### 回答\n"
+      inputs = tokenizer([prompt], return_tensors="pt").to(model.device)
+  
+      with torch.no_grad():
+          outputs = model.generate(
+              **inputs,
+              max_new_tokens=512,
+              use_cache=True,
+              do_sample=False,
+              repetition_penalty=1.2,
+          )
+      return tokenizer.decode(outputs[0], skip_special_tokens=True).split("\n### 回答")[
+          -1
+      ]
+  
+  
+  def save_results(results: List[Dict[str, Any]], output_path: str) -> None:
+      """結果をJSONLファイルに保存"""
+      output_path = Path(output_path)
+      output_path.parent.mkdir(parents=True, exist_ok=True)
+  
+      with open(output_path, "w", encoding="utf-8") as f:
+          for result in results:
+              json.dump(result, f, ensure_ascii=False)
+              f.write("\n")
+  
+  
+  def main():
+      config = ModelConfig()
+      model, tokenizer = load_model(config)
+  
+      datasets = load_datasets("./elyza-tasks-100-TV_0.jsonl")
+  
+      results = []
+      for dt in tqdm(datasets, desc="推論実行中"):
+          prediction = generate_prediction(model, tokenizer, dt["input"])
+          results.append(
+              {"task_id": dt["task_id"], "input": dt["input"], "output": prediction}
+          )
+  
+      model_basename = config.model_name.split("/")[-1]
+      save_results(results, f"/content/{model_basename}_output.jsonl")
+  
+  
+  if __name__ == "__main__":
+      main()
+
+  
+  ```