Sparkles-AI/gpt4mini
0
1import argparse2import os3import random4 5import numpy as np6import torch7import torch.backends.cudnn as cudnn8import gradio as gr9 10from minigpt4.common.config import Config11from minigpt4.common.dist_utils import get_rank12from minigpt4.common.registry import registry13from minigpt4.conversation.conversation import Chat, CONV_VISION14 15# imports modules for registration16from minigpt4.datasets.builders import *17from minigpt4.models import *18from minigpt4.processors import *19from minigpt4.runners import *20from minigpt4.tasks import *21 22def parse_args():23 parser = argparse.ArgumentParser(description="Demo")24 parser.add_argument("--cfg-path", type=str, default='eval_configs/minigpt4.yaml', help="path to configuration file.")25 parser.add_argument(26 "--options",27 nargs="+",28 help="override some settings in the used config, the key-value pair "29 "in xxx=yyy format will be merged into config file (deprecate), "30 "change to --cfg-options instead.",31 )32 args = parser.parse_args()33 return args34 35 36def setup_seeds(config):37 seed = config.run_cfg.seed + get_rank()38 39 random.seed(seed)40 np.random.seed(seed)41 torch.manual_seed(seed)42 43 cudnn.benchmark = False44 cudnn.deterministic = True45 46# ========================================47# Model Initialization48# ========================================49 50SHARED_UI_WARNING = f'''### [NOTE] It is possible that you are waiting in a lengthy queue.51 52You can duplicate and use it with a paid private GPU.53 54<a class="duplicate-button" style="display:inline-block" target="_blank" href="https://huggingface.co/spaces/Vision-CAIR/minigpt4?duplicate=true"><img style="margin-top:0;margin-bottom:0" src="https://huggingface.co/datasets/huggingface/badges/raw/main/duplicate-this-space-xl-dark.svg" alt="Duplicate Space"></a>55 56Alternatively, you can also use the demo on our [project page](https://minigpt-4.github.io).57'''58 59print('Initializing Chat')60cfg = Config(parse_args())61 62model_config = cfg.model_cfg63model_cls = registry.get_model_class(model_config.arch)64model = model_cls.from_config(model_config).to('cuda:0')65 66vis_processor_cfg = cfg.datasets_cfg.cc_align.vis_processor.train67vis_processor = registry.get_processor_class(vis_processor_cfg.name).from_config(vis_processor_cfg)68chat = Chat(model, vis_processor)69print('Initialization Finished')70 71# ========================================72# Gradio Setting73# ========================================74 75def gradio_reset(chat_state, img_list):76 if chat_state is not None:77 chat_state.messages = []78 if img_list is not None:79 img_list = []80 return None, gr.update(value=None, interactive=True), gr.update(placeholder='Please upload your image first', interactive=False), gr.update(value="Upload & Start Chat", interactive=True), chat_state, img_list81 82def upload_img(gr_img, text_input, chat_state):83 if gr_img is None:84 return None, None, gr.update(interactive=True), chat_state, None85 chat_state = CONV_VISION.copy()86 img_list = []87 llm_message = chat.upload_img(gr_img, chat_state, img_list)88 return gr.update(interactive=False), gr.update(interactive=True, placeholder='Type and press Enter'), gr.update(value="Start Chatting", interactive=False), chat_state, img_list89 90def gradio_ask(user_message, chatbot, chat_state):91 if len(user_message) == 0:92 return gr.update(interactive=True, placeholder='Input should not be empty!'), chatbot, chat_state93 chat.ask(user_message, chat_state)94 chatbot = chatbot + [[user_message, None]]95 return '', chatbot, chat_state96 97 98def gradio_answer(chatbot, chat_state, img_list, num_beams, temperature):99 llm_message = chat.answer(conv=chat_state, img_list=img_list, max_new_tokens=300, num_beams=1, temperature=temperature, max_length=2000)[0]100 chatbot[-1][1] = llm_message101 return chatbot, chat_state, img_list102 103title = """<h1 align="center">Demo of MiniGPT-4</h1>"""104description = """<h3>This is the demo of MiniGPT-4. Upload your images and start chatting!</h3>"""105article = """<div style='display:flex; gap: 0.25rem; '><a href='https://minigpt-4.github.io'><img src='https://img.shields.io/badge/Project-Page-Green'></a><a href='https://github.com/Vision-CAIR/MiniGPT-4'><img src='https://img.shields.io/badge/Github-Code-blue'></a><a href='https://github.com/TsuTikgiau/blip2-llm/blob/release_prepare/MiniGPT_4.pdf'><img src='https://img.shields.io/badge/Paper-PDF-red'></a></div>106"""107 108#TODO show examples below109 110with gr.Blocks() as demo:111 gr.Markdown(title)112 gr.Markdown(SHARED_UI_WARNING)113 gr.Markdown(description)114 gr.Markdown(article)115 116 with gr.Row():117 with gr.Column(scale=0.5):118 image = gr.Image(type="pil")119 upload_button = gr.Button(value="Upload & Start Chat", interactive=True, variant="primary")120 clear = gr.Button("Restart")121 122 num_beams = gr.Slider(123 minimum=1,124 maximum=5,125 value=1,126 step=1,127 interactive=True,128 label="beam search numbers)",129 )130 131 temperature = gr.Slider(132 minimum=0.1,133 maximum=2.0,134 value=1.0,135 step=0.1,136 interactive=True,137 label="Temperature",138 )139 140 141 with gr.Column():142 chat_state = gr.State()143 img_list = gr.State()144 chatbot = gr.Chatbot(label='MiniGPT-4')145 text_input = gr.Textbox(label='User', placeholder='Please upload your image first', interactive=False)146 147 upload_button.click(upload_img, [image, text_input, chat_state], [image, text_input, upload_button, chat_state, img_list])148 149 text_input.submit(gradio_ask, [text_input, chatbot, chat_state], [text_input, chatbot, chat_state]).then(150 gradio_answer, [chatbot, chat_state, img_list, num_beams, temperature], [chatbot, chat_state, img_list]151 )152 clear.click(gradio_reset, [chat_state, img_list], [chatbot, image, text_input, upload_button, chat_state, img_list], queue=False)153 154demo.launch(enable_queue=True)