Update app.py

app.py

@@ -2,49 +2,36 @@ import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 
-# --- NAMA MODEL YANG BARU KAMU UPLOAD TADI ---
 model_id = "forti2026/gemma-3-1b-chatbot-skripsi"
 
-print(f"Sedang mendownload model: {model_id}")
+print(f"Sedang mendownload model baru: {model_id}")
 
-# Load Tokenizer
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 
-# Load Model (CPU Mode)
-# Kita pakai float32 karena CPU Free Tier kadang error kalau dipaksa float16/4bit
-# Di app.py
 model = AutoModelForCausalLM.from_pretrained(
     model_id,
-    torch_dtype=torch.float32,
+    torch_dtype=torch.float32, 
     low_cpu_mem_usage=True
 )
 
 def chat_logic(message):
-    # Format Prompt Khusus Gemma (PENTING!)
-    # Ini supaya model tau mana user mana sistem
     input_text = f"<start_of_turn>user\n{message}<end_of_turn>\n<start_of_turn>model\n"
     
-    # Ubah teks jadi angka
     inputs = tokenizer(input_text, return_tensors="pt")
     
-    # Proses Generate Jawaban
+    # Generate
     outputs = model.generate(
         **inputs, 
-        max_new_tokens=250, # Maksimal panjang jawaban
-        do_sample=True,     # Biar jawaban bervariasi dikit (kreatif)
-        temperature=0.7,    # Tingkat 'kreativitas' (0.7 itu pas)
+        max_new_tokens=250,
+        do_sample=True,
+        temperature=0.7,
         top_k=50,
         top_p=0.95
     )
     
-    # Terjemahkan angka balik ke teks
     response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    
-    # Ambil hanya bagian jawaban si model (buang prompt user di atasnya)
     clean_response = response.split("model\n")[-1].strip()
-    
     return clean_response
 
-# Bikin Interface API
 iface = gr.Interface(fn=chat_logic, inputs="text", outputs="text")
 iface.launch()