Update

app.py

@@ -1,8 +1,7 @@
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
+from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 import os
-from threading import Thread
 
 # Nonaktifkan cache Hugging Face untuk hemat penyimpanan
 os.environ["HF_HUB_DISABLE_CACHE"] = "1"
@@ -21,7 +20,7 @@ model = AutoModelForCausalLM.from_pretrained(
 # Fungsi untuk menghasilkan respons
 def generate_response(user_input, chat_history):
     if not user_input.strip():
-        return [("Error", "Masukkan teks tidak boleh kosong!")], chat_history
+        return [{"role": "assistant", "content": "Masukkan teks tidak boleh kosong!"}], chat_history
 
     if not chat_history:
         chat_history = []
@@ -41,39 +40,29 @@ def generate_response(user_input, chat_history):
     # Tokenisasi input
     inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=False).to("cpu")
 
-    # Gunakan TextStreamer untuk streaming respons (meningkatkan UX)
-    streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
-
-    # Generate respons di thread terpisah untuk responsivitas
-    def generate():
-        outputs = model.generate(
-            **inputs,
-            max_new_tokens=200,  # Batasi token untuk kecepatan
-            do_sample=True,
-            temperature=0.75,
-            top_p=0.85,
-            eos_token_id=tokenizer.eos_token_id,
-            use_cache=True,      # Cache untuk inferensi lebih cepat
-            streamer=streamer
-        )
-        return outputs
-
-    # Jalankan generasi di thread
-    thread = Thread(target=generate)
-    thread.start()
-    thread.join()
-
-    # Ambil respons dari output streamer (decode manual)
-    bot_response = tokenizer.decode(
-        model.generate(**inputs, max_new_tokens=200, do_sample=True, temperature=0.75, top_p=0.85)[0][inputs.input_ids.shape[-1]:],
-        skip_special_tokens=True
+    # Generate respons
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=200,  # Batasi token untuk kecepatan
+        do_sample=True,
+        temperature=0.75,
+        top_p=0.85,
+        eos_token_id=tokenizer.eos_token_id,
+        use_cache=True       # Cache untuk inferensi lebih cepat
     )
 
+    # Decode respons
+    bot_response = tokenizer.decode(outputs[0][inputs.input_ids.shape[-1]:], skip_special_tokens=True)
+
     # Perbarui riwayat percakapan
     chat_history.append((user_input, bot_response))
     
-    # Format output untuk Gradio Chatbot
-    return [(user_msg, bot_msg) for user_msg, bot_msg in chat_history], chat_history
+    # Format output untuk Gradio Chatbot (format messages)
+    return [
+        {"role": "user" if i % 2 == 0 else "assistant", "content": msg}
+        for i, (user_msg, bot_msg) in enumerate(chat_history)
+        for msg in [user_msg, bot_msg]
+    ], chat_history
 
 # Fungsi untuk menghapus riwayat
 def clear_history():
@@ -83,7 +72,7 @@ def clear_history():
 with gr.Blocks(
     theme=gr.themes.Monochrome(),  # Tema modern dan bersih
     css="""
-        #chatbot {border-radius: 10px; border: 1px solid #e0e0e0;}
+        #chatbot {border-radius: 10px; border: 1px solid #e0e0e0; padding: 10px;}
         .gradio-container {max-width: 800px; margin: auto;}
         #input-box {border-radius: 8px;}
         #submit-btn, #clear-btn {border-radius: 8px; background: #007bff; color: white;}
@@ -99,11 +88,10 @@ with gr.Blocks(
     
     # Komponen UI
     chatbot = gr.Chatbot(
-        label="Percakapan",
+        type="messages",  # Gunakan format messages untuk kompatibilitas
         height=450,
         show_label=False,
-        elem_id="chatbot",
-        bubble_full_width=False
+        elem_id="chatbot"
     )
     with gr.Row():
         user_input = gr.Textbox(
@@ -123,8 +111,7 @@ with gr.Blocks(
     submit_button.click(
         fn=generate_response,
         inputs=[user_input, chat_history],
-        outputs=[chatbot, chat_history],
-        _js="() => {document.querySelector('input').value = '';}"  # Kosongkan input
+        outputs=[chatbot, chat_history]
     )
     clear_button.click(
         fn=clear_history,