update app

app.py

@@ -5,8 +5,19 @@ import torch, os, uvicorn
 
 app = FastAPI()
 
-model_name = "Qwen/Qwen-1_8B-Chat"  # ganti sesuai ukuran
+model_name = "Qwen/Qwen-1_8B-Chat"
+
 tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
+# Pasang template manual kalau tidak tersedia
+if tokenizer.chat_template is None:
+    tokenizer.chat_template = """{% for message in messages %}
+{% if message['role'] == 'system' %}{{ '<|system|>\n' + message['content'] + '\n' }}
+{% elif message['role'] == 'user' %}{{ '<|user|>\n' + message['content'] + '\n' }}
+{% elif message['role'] == 'assistant' %}{{ '<|assistant|>\n' + message['content'] + '\n' }}
+{% endif %}
+{% endfor %}<|assistant|>
+"""
+
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
     trust_remote_code=True,
@@ -20,15 +31,14 @@ class ChatRequest(BaseModel):
 
 @app.post("/chat")
 def chat(req: ChatRequest):
-    # Format percakapan sesuai template Qwen
     messages = [
         {"role": "system", "content": "You are a helpful AI assistant."},
         {"role": "user", "content": req.prompt},
     ]
 
     text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-
     inputs = tokenizer(text, return_tensors="pt").to(model.device)
+
     outputs = model.generate(**inputs, max_new_tokens=req.max_new_tokens)
     reply = tokenizer.decode(outputs[0], skip_special_tokens=True)