Update README.md

README.md

@@ -31,6 +31,74 @@ LLM-jp-3-13bに対して以下のデータセットを用いてSFTを行った
 サンプルコード(ipynb)がレポジトリに含まれています。
 `dakesan0-inference-testcode.ipynb`
 
+unslothを用いた推論のみ動作を確認しています。
+
+```py
+from unsloth import FastLanguageModel
+from peft import PeftModel
+import torch
+import json
+from tqdm import tqdm
+import re
+import datasets
+
+model_id = "llm-jp/llm-jp-3-13b"
+adapter_id = "poprap/llm-jp-3-13b-it-2-3"
+adapter_dpo_id = "poprap/llm-jp-3-13b-dpo"
+
+dtype = None 
+load_in_4bit = True 
+
+model, tokenizer = FastLanguageModel.from_pretrained(
+    model_name=model_id,
+    dtype=dtype,
+    load_in_4bit=load_in_4bit,
+    trust_remote_code=True,
+)
+
+model = PeftModel.from_pretrained(model, adapter_id, token = HF_TOKEN)
+model = PeftModel.from_pretrained(model, adapter_dpo_id, token = HF_TOKEN)
+
+ds = []
+
+with open("elyza-tasks-100-TV_0.jsonl", "r") as f:
+    item = ""
+    for line in f:
+        line = line.strip()
+        item += line
+        if item.endswith("}"):
+            ds.append(json.loads(item))
+            item = ""
+
+# 推論するためにモデルのモードを変更
+FastLanguageModel.for_inference(model)
+
+results = []
+for dt in tqdm(ds):
+    input = dt["input"]
+
+    prompt = f"""### 指示\n{input}\n上記指示に簡潔に回答してください。\n### 回答\n"""
+
+    inputs = tokenizer([prompt], return_tensors = "pt").to(model.device)
+
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=1024,
+        use_cache = True, 
+        do_sample=False, 
+        repetition_penalty=1.2
+      )
+    prediction = tokenizer.decode(outputs[0], skip_special_tokens=True).split('\n### 回答')[-1]
+    
+    results.append({"task_id": dt['task_id'], "input": input, "output": prediction})
+
+json_file_id = re.sub(".*/", "", adapter_id)
+with open(f"{json_file_id}_output.jsonl", 'w', encoding='utf-8') as f:
+    for result in results:
+        json.dump(result, f, ensure_ascii=False)
+        f.write('\n')
+```
+
 # Special thanks
 
 本コンペを運営いただいた方々に深く御礼申し上げます。