Update app.py

app.py

@@ -25,44 +25,44 @@ hf_hub_download(
 # 推論関数
 @spaces.GPU(duration=120)
 def respond(
-   message,
-   history: list[tuple[str, str]],
-   model,
-   system_message,
-   max_tokens,
-   temperature,
-   top_p,
-   top_k,
-   repeat_penalty,
-):
-   chat_template = MessagesFormatterType.GEMMA_2
-
-   llm = Llama(
+    message,
+    history: list[tuple[str, str]],
+    model,
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+    top_k,
+    repeat_penalty,
+    ):
+    chat_template = MessagesFormatterType.GEMMA_2
+    
+    llm = Llama(
        model_path=f"models/{model}",
        flash_attn=True,
        n_gpu_layers=81,
        n_batch=1024,
        n_ctx=8192,
-   )
-   provider = LlamaCppPythonProvider(llm)
-
-   agent = LlamaCppAgent(
+    )
+    provider = LlamaCppPythonProvider(llm)
+    
+    agent = LlamaCppAgent(
        provider,
        system_prompt=f"{system_message}",
        predefined_messages_formatter_type=chat_template,
        debug_output=True
-   )
-
-   settings = provider.get_provider_default_settings()
-   settings.temperature = temperature
-   settings.top_k = top_k
-   settings.top_p = top_p
-   settings.max_tokens = max_tokens
-   settings.repeat_penalty = repeat_penalty
-   settings.stream = True
-
-   messages = BasicChatHistory()
-
+    )
+
+    settings = provider.get_provider_default_settings()
+    settings.temperature = temperature
+    settings.top_k = top_k
+    settings.top_p = top_p
+    settings.max_tokens = max_tokens
+    settings.repeat_penalty = repeat_penalty
+    settings.stream = True
+    
+    messages = BasicChatHistory()
+    
     one_shot_prompt = """
     あなたは優れた翻訳者です。以下の文章を日本語から英語に翻訳してください。翻訳は文法的に正しく、自然な表現を使用し、文脈に適した内容にしてください。また、専門用語や文化的なニュアンスを正確に伝えるよう心がけてください。
     
@@ -81,9 +81,8 @@ def respond(
     以下の日本語の文章を英語に翻訳してください:
     """
     system_message += one_shot_prompt
-
-
-   for msn in history:
+    
+    for msn in history:
        user = {
            'role': Roles.user,
            'content': msn[0]
@@ -94,17 +93,17 @@ def respond(
        }
        messages.add_message(user)
        messages.add_message(assistant)
-
-   stream = agent.get_chat_response(
+    
+    stream = agent.get_chat_response(
        message,
        llm_sampling_settings=settings,
        chat_history=messages,
        returns_streaming_generator=True,
        print_output=False
-   )
-
-   outputs = ""
-   for output in stream:
+    )
+    
+    outputs = ""
+    for output in stream:
        outputs += output
        yield outputs