Update app.py

app.py

@@ -1,24 +1,30 @@
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+import torch
 import gradio as gr
-import json
 
-# Eğitim verisini yükle
-with open("memory_questions.json", "r") as f:
-    memory_data = json.load(f)
+# LLaMA 2 Chat modeli
+model_id = "meta-llama/Llama-2-7b-chat-hf"
 
-# Soru tahmini fonksiyonu
-def generate_question(memory: str):
-    # En yakın eşleşmeyi manuel kurarak (örnek amaçlı)
-    for item in memory_data:
-        if item['description'].lower() in memory.lower():
-            return item['question']
-    return "What question does this memory bring to your mind?"
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    torch_dtype=torch.float16,  # CPU çalışıyorsan float32 olabilir
+    device_map="auto"
+)
+
+def generate_question(memory):
+    prompt = f"[INST] You are a helpful assistant. Based on this memory, generate a question that would help the user recall more details:\n\nMemory: {memory}\n\nQuestion: [/INST]"
+    inputs = tokenizer(prompt, return_tensors="pt")
+    outputs = model.generate(**inputs, max_new_tokens=50)
+    result = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    return result.split("Question:")[-1].strip()
 
-# Gradio UI
+# Arayüz
 iface = gr.Interface(
     fn=generate_question,
     inputs=gr.Textbox(label="Your Memory"),
     outputs=gr.Textbox(label="Generated Question"),
-    title="MemoRease Question Generator"
+    title="LLaMA Chat Question Generator"
 )
 
 iface.launch()