import gradio as gr import spaces #from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor # from qwen_vl_utils import process_vision_info import torch import base64 from PIL import Image, ImageDraw from io import BytesIO import re from deepseek_vl2.models import DeepseekVLV2Processor, DeepseekVLV2ForCausalLM from deepseek_vl2.utils.io import load_pil_images from transformers import AutoModelForCausalLM models = { "OS-Copilot/OS-Atlas-Base-7B": Qwen2VLForConditionalGeneration.from_pretrained("OS-Copilot/OS-Atlas-Base-7B", torch_dtype="auto", device_map="auto"), } processors = { "OS-Copilot/OS-Atlas-Base-7B": AutoProcessor.from_pretrained("OS-Copilot/OS-Atlas-Base-7B") } def image_to_base64(image): buffered = BytesIO() image.save(buffered, format="PNG") img_str = base64.b64encode(buffered.getvalue()).decode("utf-8") return img_str def draw_bounding_boxes(image, bounding_boxes, outline_color="red", line_width=2): draw = ImageDraw.Draw(image) for box in bounding_boxes: xmin, ymin, xmax, ymax = box draw.rectangle([xmin, ymin, xmax, ymax], outline=outline_color, width=line_width) return image def rescale_bounding_boxes(bounding_boxes, original_width, original_height, scaled_width=1000, scaled_height=1000): x_scale = original_width / scaled_width y_scale = original_height / scaled_height rescaled_boxes = [] for box in bounding_boxes: xmin, ymin, xmax, ymax = box rescaled_box = [ xmin * x_scale, ymin * y_scale, xmax * x_scale, ymax * y_scale ] rescaled_boxes.append(rescaled_box) return rescaled_boxes def deepseek(): print("helloe!!!!") # specify the path to the model model_path = "deepseek-ai/deepseek-vl2-tiny" vl_chat_processor: DeepseekVLV2Processor = DeepseekVLV2Processor.from_pretrained(model_path) tokenizer = vl_chat_processor.tokenizer vl_gpt: DeepseekVLV2ForCausalLM = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True) vl_gpt = vl_gpt.to(torch.bfloat16).cuda().eval() ## single image conversation example conversation = [ { "role": "<|User|>", "content": "\n<|ref|>The giraffe at the back.<|/ref|>.", "images": ["./images/visual_grounding_1.jpeg"], }, {"role": "<|Assistant|>", "content": ""}, ] # load images and prepare for inputs pil_images = load_pil_images(conversation) prepare_inputs = vl_chat_processor( conversations=conversation, images=pil_images, force_batchify=True, system_prompt="" ).to(vl_gpt.device) # run image encoder to get the image embeddings inputs_embeds = vl_gpt.prepare_inputs_embeds(**prepare_inputs) # run the model to get the response outputs = vl_gpt.language.generate( inputs_embeds=inputs_embeds, attention_mask=prepare_inputs.attention_mask, pad_token_id=tokenizer.eos_token_id, bos_token_id=tokenizer.bos_token_id, eos_token_id=tokenizer.eos_token_id, max_new_tokens=512, do_sample=False, use_cache=True ) answer = tokenizer.decode(outputs[0].cpu().tolist(), skip_special_tokens=False) print(f"{prepare_inputs['sft_format'][0]}", answer) @spaces.GPU def run_example(image, text_input, model_id="OS-Copilot/OS-Atlas-Base-7B"): deepseek() def run_example_old(image, text_input, model_id="OS-Copilot/OS-Atlas-Base-7B"): model = models[model_id].eval() processor = processors[model_id] prompt = f"In this UI screenshot, what is the position of the element corresponding to the command \"{text_input}\" (with bbox)?" messages = [ { "role": "user", "content": [ {"type": "image", "image": f"data:image;base64,{image_to_base64(image)}"}, {"type": "text", "text": prompt}, ], } ] text = processor.apply_chat_template( messages, tokenize=False, add_generation_prompt=True ) image_inputs, video_inputs = process_vision_info(messages) inputs = processor( text=[text], images=image_inputs, videos=video_inputs, padding=True, return_tensors="pt", ) inputs = inputs.to("cuda") generated_ids = model.generate(**inputs, max_new_tokens=128) generated_ids_trimmed = [ out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids) ] output_text = processor.batch_decode( generated_ids_trimmed, skip_special_tokens=False, clean_up_tokenization_spaces=False ) print(output_text) text = output_text[0] object_ref_pattern = r"<\|object_ref_start\|>(.*?)<\|object_ref_end\|>" box_pattern = r"<\|box_start\|>(.*?)<\|box_end\|>" object_ref = re.search(object_ref_pattern, text).group(1) box_content = re.search(box_pattern, text).group(1) boxes = [tuple(map(int, pair.strip("()").split(','))) for pair in box_content.split("),(")] boxes = [[boxes[0][0], boxes[0][1], boxes[1][0], boxes[1][1]]] scaled_boxes = rescale_bounding_boxes(boxes, image.width, image.height) return object_ref, scaled_boxes, draw_bounding_boxes(image, scaled_boxes) css = """ #output { height: 500px; overflow: auto; border: 1px solid #ccc; } """ with gr.Blocks(css=css) as demo: gr.Markdown( """ # Demo for OS-ATLAS: A Foundation Action Model For Generalist GUI Agents """) with gr.Row(): with gr.Column(): input_img = gr.Image(label="Input Image", type="pil") model_selector = gr.Dropdown(choices=list(models.keys()), label="Model", value="OS-Copilot/OS-Atlas-Base-7B") text_input = gr.Textbox(label="User Prompt") submit_btn = gr.Button(value="Submit") with gr.Column(): model_output_text = gr.Textbox(label="Model Output Text") model_output_box = gr.Textbox(label="Model Output Box") annotated_image = gr.Image(label="Annotated Image") gr.Examples( examples=[ ["assets/web_6f93090a-81f6-489e-bb35-1a2838b18c01.png", "select search textfield"], ["assets/web_6f93090a-81f6-489e-bb35-1a2838b18c01.png", "switch to discussions"], ], inputs=[input_img, text_input], outputs=[model_output_text, model_output_box, annotated_image], fn=run_example, cache_examples=True, label="Try examples" ) submit_btn.click(run_example, [input_img, text_input, model_selector], [model_output_text, model_output_box, annotated_image]) demo.launch(debug=True)