Update app.py

app.py

@@ -1,17 +1,21 @@
 import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer
+from peft import PeftModel  # Pastikan PEFT terinstal
 
-# Load the model and tokenizer from Hugging Face Hub
-model_name = "nicolaakmal/llama32-lora-finetuned-v2"
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name)
+# Tentukan model dasar dan LoRA weights
+base_model_name = "meta-llama/Llama-3.2-3B-Instruct"  # Model dasar sesuai yang ada di gambar
+lora_model_name = "nicolaakmal/llama32-lora-finetuned-v3"
 
-# Function to generate a response from the model
+# Load tokenizer dan base model
+tokenizer = AutoTokenizer.from_pretrained(base_model_name)
+base_model = AutoModelForCausalLM.from_pretrained(base_model_name)
+
+# Load LoRA model sebagai PeftModel
+model = PeftModel.from_pretrained(base_model, lora_model_name)
+
+# Fungsi untuk menghasilkan respons
 def generate_response(prompt):
-    # Mengatur input untuk model menggunakan tokenizer
     inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
-    
-    # Menggunakan TextStreamer untuk streaming teks keluaran
     text_streamer = TextStreamer(tokenizer, skip_prompt=True)
     outputs = model.generate(
         input_ids=inputs["input_ids"], 
@@ -21,12 +25,10 @@ def generate_response(prompt):
         temperature=1.5, 
         top_p=0.1
     )
-    
-    # Menghasilkan respons dari streamer
     response = tokenizer.decode(outputs[0], skip_special_tokens=True)
     return response
 
-# Create Gradio interface
+# Interface Gradio
 iface = gr.Interface(
     fn=generate_response,
     inputs="text",