Update app.py

app.py

@@ -2,18 +2,58 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 import gradio as gr
 import torch
 
-device = "cuda" if torch.cuda.is_available() else "cpu"
-model = AutoModelForCausalLM.from_pretrained("naver-hyperclovax/HyperCLOVAX-SEED-Text-Instruct-0.5B").to(device)
-tokenizer = AutoTokenizer.from_pretrained("naver-hyperclovax/HyperCLOVAX-SEED-Text-Instruct-0.5B")
+# CPU 강제 설정
+device = "cpu"
+torch.set_num_threads(4)  # CPU 스레드 제한
+
+# 경량화된 모델 로드
+model = AutoModelForCausalLM.from_pretrained(
+    "naver-hyperclovax/HyperCLOVAX-SEED-Text-Instruct-0.5B",
+    torch_dtype=torch.float32,  # CPU는 float32 권장
+    low_cpu_mem_usage=True
+).to(device)
+
+tokenizer = AutoTokenizer.from_pretrained(
+    "naver-hyperclovax/HyperCLOVAX-SEED-Text-Instruct-0.5B"
+)
 
 def predict(message, history):
+    # 메모리 절약을 위한 간소화된 채팅 구성
     chat = [
-        {"role": "system", "content": "당신은 CLOVA X AI입니다."},
-        *[{"role": "user" if i%2==0 else "assistant", "content": h} for i,h in enumerate(sum(history, ()))],
+        {"role": "system", "content": "간결하게 답변해주세요."},
         {"role": "user", "content": message}
     ]
-    inputs = tokenizer.apply_chat_template(chat, return_tensors="pt").to(device)
-    outputs = model.generate(inputs, max_length=1024, temperature=0.7)
-    return tokenizer.decode(outputs[0][inputs.shape[1]:], skip_special_tokens=True)
+    
+    # CPU 최적화 설정
+    inputs = tokenizer.apply_chat_template(
+        chat,
+        return_tensors="pt",
+        max_length=512,  # 길이 제한
+        truncation=True
+    ).to(device)
+    
+    outputs = model.generate(
+        inputs,
+        max_new_tokens=200,  # 짧은 응답
+        temperature=0.3,    # 창의성 감소
+        do_sample=False      #确定性 응답
+    )
+    
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+# 경량화된 인터페이스
+demo = gr.ChatInterface(
+    predict,
+    title="CLOVA X (CPU 모드)",
+    description="CPU 전용 경량화 버전",
+    theme="soft",
+    examples=["안녕하세요", "날씨 알려줘"]
+)
 
-gr.ChatInterface(predict).launch(server_name="0.0.0.0")
+if __name__ == "__main__":
+    demo.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        favicon_path=None,
+        prevent_thread_lock=True
+    )