Update app.py

app.py

@@ -1,101 +1,58 @@
+from unsloth import FastLanguageModel
 import torch
 import gradio as gr
-import os
-from threading import Thread
-from typing import Iterator
-from transformers import (
-    AutoModelForCausalLM,
-    BitsAndBytesConfig,
-    GenerationConfig,
-    AutoTokenizer,
-    TextIteratorStreamer,
-)
-from peft import AutoPeftModelForCausalLM
 
-
-
-#deklarasi
 max_seq_length = 2048 # Choose any! We auto support RoPE Scaling internally!
 dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
 load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False.
-#alpaca_prompt = """Berikut adalah instruksi yang deskripsikan tugas dan sepasang input dan konteksnya. Tulis response sesuai dengan permintaan.
+alpaca_prompt = """Berikut adalah instruksi yang deskripsikan tugas dan sepasang input dan konteksnya. Tulis response sesuai dengan permintaan.
 ### Instruction:
 {}
 ### Input:
 {}
 ### Response:
-#{}"""
-
-device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
-MAX_MAX_NEW_TOKENS = 2048
-DEFAULT_MAX_NEW_TOKENS = 1024
-MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
-
-model_id = "abdfajar707/llama3_8B_lora_model_rkp_pn2025_v3"
-#tokenizer = LlamaTokenizer.from_pretrained(model_id)
-#model, tokenizer = AutoModelForCausalLM.from_pretrained(
-#    model_id,
-#    device_map="auto",
-#   quantization_config=BitsAndBytesConfig(load_in_8bit=True),
-#)
-model = AutoPeftModelForCausalLM.from_pretrained(
-        model_id, # YOUR MODEL YOU USED FOR TRAINING
+{}"""
+
+if True:
+    from unsloth import FastLanguageModel
+    model, tokenizer = FastLanguageModel.from_pretrained(
+        model_name = "abdfajar707/llama3_8B_lora_model_orpo_v1", # YOUR MODEL YOU USED FOR TRAINING
+        max_seq_length = max_seq_length,
+        dtype = dtype,
         load_in_4bit = load_in_4bit,
     )
-tokenizer = AutoTokenizer.from_pretrained(
-            model_id,
-            quantization_config=BitsAndBytesConfig(load_in_8bit=True)
-    )
-
-model.config.sliding_window = 4096
-model.eval()
-
+    FastLanguageModel.for_inference(model) # Enable native 2x faster inference
 
-#@spaces.GPU(duration=90)
-def generate(
-    message: str,
-    chat_history: list[tuple[str, str]],
-    max_new_tokens: int = 1024,
-    temperature: float = 0.6,
-    top_p: float = 0.9,
-    top_k: int = 50,
-    repetition_penalty: float = 1.2,
-) -> Iterator[str]:
-    conversation = []
-    for user, assistant in chat_history:
-        conversation.extend(
-            [
-                {"role": "user", "content": user},
-                {"role": "assistant", "content": assistant},
-            ]
-        )
-    conversation.append({"role": "user", "content": message})
 
-    input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt")
-    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
-        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
-        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
-    input_ids = input_ids.to(model.device)
 
-    streamer = TextIteratorStreamer(tokenizer, timeout=20.0, skip_prompt=True, skip_special_tokens=True)
-    generate_kwargs = dict(
-        {"input_ids": input_ids},
-        streamer=streamer,
-        max_new_tokens=max_new_tokens,
-        do_sample=True,
-        top_p=top_p,
-        top_k=top_k,
-        temperature=temperature,
-        num_beams=1,
-        repetition_penalty=repetition_penalty,
+# Fungsi untuk menghasilkan respons
+def generate_response(prompt, max_length=1000):
+    inputs = tokenizer(
+[
+    alpaca_prompt.format(
+        prompt, # instruction
+        "", # input
+        "", # output - leave this blank for generation!
     )
-    t = Thread(target=model.generate, kwargs=generate_kwargs)
-    t.start()
+], return_tensors = "pt").to("cuda")
+    outputs = model.generate(**inputs, max_length=max_length, pad_token_id=tokenizer.eos_token_id)
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    return response
+
+history = []
+def wrapper_chat_history(chat_history, history):
+    chat_history = history[1:]
+    return chat_history
+
+def converse(message, chat_history):
+    response = generate_response(message)
+    print(response)
+    user_msg = {"role": "user", "content": message}
+    history.append(user_msg)
+    ai_msg = {"role": "assistant", "content": response}
+    history.append(ai_msg)
+    return history[-1]["content"]
 
-    outputs = []
-    for text in streamer:
-        outputs.append(text)
-        yield "".join(outputs)
 
 DESCRIPTION = '''
 <div style="padding: 5px; text-align: left; display: flex; flex-direction: column; align-items: left;">
@@ -131,64 +88,21 @@ h1 {
   border-radius: 100vh;
 }
 """
-chatbot=gr.Chatbot(height=450, placeholder=PLACEHOLDER, label='Interlinked Sytem ChatInterface')
-
-
-chat_interface = gr.ChatInterface(
-    fn=generate,
-    chatbot=chatbot,
-    additional_inputs=[
-        gr.Slider(
-            label="Max new tokens",
-            minimum=1,
-            maximum=MAX_MAX_NEW_TOKENS,
-            step=1,
-            value=DEFAULT_MAX_NEW_TOKENS,
-        ),
-        gr.Slider(
-            label="Temperature",
-            minimum=0.1,
-            maximum=4.0,
-            step=0.1,
-            value=0.6,
-        ),
-        gr.Slider(
-            label="Top-p (nucleus sampling)",
-            minimum=0.05,
-            maximum=1.0,
-            step=0.05,
-            value=0.9,
-        ),
-        gr.Slider(
-            label="Top-k",
-            minimum=1,
-            maximum=1000,
-            step=1,
-            value=50,
-        ),
-        gr.Slider(
-            label="Repetition penalty",
-            minimum=1.0,
-            maximum=2.0,
-            step=0.05,
-            value=1.2,
-        ),
-    ],
-    stop_btn=None,
-    examples=[
-        ["Apa yang dimaksud dengan RPJMN"],
-        ["Jelaskan tentang RPJMN 2020-2024"],
-        ["Apa peran RKP 2021 dan 20211 dalam RPJM 2020-2024"],
-        ["Apa saja program prioritas RPJMN 2020-2024"],
-    ],
-)
-
-
-
-with gr.Blocks(css=css, fill_height=True) as demo:
-    gr.Markdown(DESCRIPTION)
-    #gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
-    chat_interface.render()
+chatbot=gr.Chatbot(height=500, placeholder=PLACEHOLDER, label='Interlinked Sytem ChatInterface')
+    
+with gr.Blocks(css=css) as interface:
+  chatbot=chatbot,
+  with gr.Row():
+    with gr.Column(scale=1):
+      gr.HTML('<img src="https://datahub.data.go.id/data/static/Kementerian%20PPN%20Bappenas%20Tanpa%20Teks.png" width="100px" alt="Image" style="max-width: 100%;">')
+  with gr.Row():
+    with gr.Column(scale=1, elem_id='col'):
+      gr.ChatInterface(fn=converse, title=("""
+<center> 
+<h1>KemenPPN/Bappenas</h1>
+<b>AI-Interlinked System/Bappenas GPT<b>
+</center>
+"""
+))
 
-if __name__ == "__main__":
-    demo.queue(max_size=20).launch()
+interface.launch()