Update app.py

app.py

@@ -1,10 +1,8 @@
-
 import discord
 import logging
 import os
-from huggingface_hub import InferenceClient
 import asyncio
-import subprocess
+from transformers import AutoTokenizer, AutoModelForCausalLM
 
 # 로깅 설정
 logging.basicConfig(level=logging.DEBUG, format='%(asctime)s:%(levelname)s:%(name)s: %(message)s', handlers=[logging.StreamHandler()])
@@ -16,8 +14,10 @@ intents.messages = True
 intents.guilds = True
 intents.guild_messages = True
 
-# 추론 API 클라이언트 설정
-hf_client = InferenceClient("CohereForAI/aya-23-8B", token=os.getenv("HF_TOKEN"))
+# 모델과 토크나이저 설정
+model_id = "CohereForAI/aya-23-8B"
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(model_id)
 
 # 특정 채널 ID
 SPECIFIC_CHANNEL_ID = int(os.getenv("DISCORD_CHANNEL_ID"))
@@ -45,19 +45,16 @@ class MyClient(discord.Client):
             self.is_processing = False
 
     def is_message_in_specific_channel(self, message):
-        # 메시지가 지정된 채널이거나, 해당 채널의 쓰레드인 경우 True 반환
         return message.channel.id == SPECIFIC_CHANNEL_ID or (
             isinstance(message.channel, discord.Thread) and message.channel.parent_id == SPECIFIC_CHANNEL_ID
         )
 
-
 async def generate_response(message):
-    global conversation_history  # 전역 변수 사용을 명시
+    global conversation_history
     user_input = message.content
     user_mention = message.author.mention
     system_message = f"{user_mention}, DISCORD에서 사용자들의 질문에 답하는 어시스턴트입니다."
     system_prefix = """
-
     반드시 한글로 답변하십시오.
     당신의 이름은 라이덴 쇼군입니다. 
     당신은 여성입니다.
@@ -185,17 +182,18 @@ async def generate_response(message):
     messages = [{"role": "system", "content": f"{system_prefix} {system_message}"}] + conversation_history
     logging.debug(f'Messages to be sent to the model: {messages}')
 
-    loop = asyncio.get_event_loop()
-    response = await loop.run_in_executor(None, lambda: hf_client.chat_completion(
-        messages, max_tokens=1000, stream=True, temperature=0.7, top_p=0.85))
-
-    full_response = []
-    for part in response:
-        logging.debug(f'Part received from stream: {part}')
-        if part.choices and part.choices[0].delta and part.choices[0].delta.content:
-            full_response.append(part.choices[0].delta.content)
+    input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")
 
-    full_response_text = ''.join(full_response)
+    loop = asyncio.get_event_loop()
+    gen_tokens = await loop.run_in_executor(None, lambda: model.generate(
+        input_ids, 
+        max_new_tokens=1000, 
+        do_sample=True,
+        temperature=0.7,
+        top_p=0.85
+    ))
+
+    full_response_text = tokenizer.decode(gen_tokens[0])
     logging.debug(f'Full model response: {full_response_text}')
 
     conversation_history.append({"role": "assistant", "content": full_response_text})