Update README.md

README.md

@@ -1,3 +1,86 @@
 ---
 license: mit
+language: ja
+    
 ---
+
+```python
+
+import torch
+from peft import PeftModel, PeftConfig
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_name = "cyberagent/open-calm-large"
+lora_weights = "Mizuiro-sakura/open-calm-large-finetuned-databricks-dolly"
+
+# モデルの準備
+model = AutoModelForCausalLM.from_pretrained(
+    model_name
+)
+
+# トークンナイザーの準備
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+# LoRAモデルの準備
+model = PeftModel.from_pretrained(
+    model, 
+    lora_weights,
+    adapter_name=lora_weights
+)
+
+# 評価モード
+model.eval()
+
+# プロンプトテンプレートの準備
+def generate_prompt(data_point):
+    if data_point["input"]:
+        return f"""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:
+{data_point["instruction"]}
+
+### Input:
+{data_point["input"]}
+
+### Response:"""
+    else:
+        return f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.
+
+### Instruction:
+{data_point["instruction"]}
+
+### Response:"""
+
+# テキスト生成関数の定義
+def generate(instruction,input=None,maxTokens=256):
+    # 推論
+    prompt = generate_prompt({'instruction':instruction,'input':input})
+    input_ids = tokenizer(prompt, return_tensors="pt", truncation=True).input_ids.to('mps')
+    outputs = model.generate(
+        input_ids=input_ids, 
+        max_new_tokens=maxTokens, 
+        do_sample=True,
+        temperature=0.7, 
+        top_p=0.75, 
+        top_k=40,         
+        no_repeat_ngram_size=2,
+    )
+    outputs = outputs[0].tolist()
+
+    # EOSトークンにヒットしたらデコード完了
+    if tokenizer.eos_token_id in outputs:
+        eos_index = outputs.index(tokenizer.eos_token_id)
+    else:
+        eos_index = len(outputs)
+    decoded = tokenizer.decode(outputs[:eos_index])
+    
+    # レスポンス内容のみ抽出
+    sentinel = "### Response:"
+    sentinelLoc = decoded.find(sentinel)
+    if sentinelLoc >= 0:
+        print(decoded[sentinelLoc+len(sentinel):])
+    else:
+        print('Warning: Expected prompt template to be emitted.  Ignoring output.')
+
+generate("自然言語処理とは？")
+```