Update app.py

app.py

@@ -9,7 +9,26 @@ app = FastAPI()
 # โหลดโมเดลและ tokenizer
 model_name = "scb10x/llama-3-typhoon-v1.5-8b-instruct"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
+
+# ตรวจสอบว่ามี GPU หรือไม่
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# โหลดโมเดลด้วยการตั้งค่าที่เหมาะสม
+if device == "cuda":
+    model = AutoModelForCausalLM.from_pretrained(
+        model_name, 
+        torch_dtype=torch.float16, 
+        device_map="auto",
+        low_cpu_mem_usage=True
+    )
+else:
+    model = AutoModelForCausalLM.from_pretrained(
+        model_name,
+        torch_dtype=torch.float32,
+        low_cpu_mem_usage=True
+    )
+
+model.to(device)
 
 class Query(BaseModel):
     queryResult: Optional[dict] = None
@@ -25,9 +44,10 @@ async def webhook(query: Query):
         
         # สร้าง prompt และ generate ข้อความ
         prompt = f"Human: {user_query}\nAI:"
-        input_ids = tokenizer.encode(prompt, return_tensors="pt").to(model.device)
+        input_ids = tokenizer.encode(prompt, return_tensors="pt").to(device)
         
-        output = model.generate(input_ids, max_new_tokens=100, temperature=0.7)
+        with torch.no_grad():
+            output = model.generate(input_ids, max_new_tokens=100, temperature=0.7)
         response = tokenizer.decode(output[0], skip_special_tokens=True)
         
         # แยกส่วนที่เป็นคำตอบของ AI
@@ -35,4 +55,8 @@ async def webhook(query: Query):
         
         return {"fulfillmentText": ai_response}
     except Exception as e:
-        raise HTTPException(status_code=500, detail=str(e))
+        raise HTTPException(status_code=500, detail=str(e))
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=7860)