import os import torch import requests import gradio as gr import transformers from transformers import Pix2StructForConditionalGeneration, Pix2StructProcessor from peft import PeftModel ## CoT prompts def _add_markup(table): parts = [p.strip() for p in table.splitlines(keepends=False)] if parts[0].startswith('TITLE'): result = f"Title: {parts[0].split(' | ')[1].strip()}\n" rows = parts[1:] else: result = '' rows = parts prefixes = ['Header: '] + [f'Row {i+1}: ' for i in range(len(rows) - 1)] return result + '\n'.join(prefix + row for prefix, row in zip(prefixes, rows)) _TABLE = """Year | Democrats | Republicans | Independents 2004 | 68.1% | 45.0% | 53.0% 2006 | 58.0% | 42.0% | 53.0% 2007 | 59.0% | 38.0% | 45.0% 2009 | 72.0% | 49.0% | 60.0% 2011 | 71.0% | 51.2% | 58.0% 2012 | 70.0% | 48.0% | 53.0% 2013 | 72.0% | 41.0% | 60.0%""" _INSTRUCTION = 'Read the table below to answer the following questions.' _TEMPLATE = f"""First read an example then the complete question for the second table. ------------ {_INSTRUCTION} {_add_markup(_TABLE)} Q: In which year republicans have the lowest favor rate? A: Let's find the column of republicans. Then let's extract the favor rates, they [45.0, 42.0, 38.0, 49.0, 51.2, 48.0, 41.0]. The smallest number is 38.0, that's Row 3. Row 3 is year 2007. The answer is 2007. Q: What is the sum of Democrats' favor rates of 2004, 2012, and 2013? A: Let's find the rows of years 2004, 2012, and 2013. We find Row 1, 6, 7. The favor dates of Demoncrats on that 3 rows are 68.1, 70.0, and 72.0. 68.1+70.0+72=210.1. The answer is 210.1. Q: By how many points do Independents surpass Republicans in the year of 2011? A: Let's find the row with year = 2011. We find Row 5. We extract Independents and Republicans' numbers. They are 58.0 and 51.2. 58.0-51.2=6.8. The answer is 6.8. Q: Which group has the overall worst performance? A: Let's sample a couple of years. In Row 1, year 2004, we find Republicans having the lowest favor rate 45.0 (since 45.0<68.1, 45.0<53.0). In year 2006, Row 2, we find Republicans having the lowest favor rate 42.0 (42.0<58.0, 42.0<53.0). The trend continues to other years. The answer is Republicans. Q: Which party has the second highest favor rates in 2007? A: Let's find the row of year 2007, that's Row 3. Let's extract the numbers on Row 3: [59.0, 38.0, 45.0]. 45.0 is the second highest. 45.0 is the number of Independents. The answer is Independents. {_INSTRUCTION}""" ## alpaca-lora assert ( "LlamaTokenizer" in transformers._import_structure["models.llama"] ), "LLaMA is now in HuggingFace's main branch.\nPlease reinstall it: pip uninstall transformers && pip install git+https://github.com/huggingface/transformers.git" from transformers import LlamaTokenizer, LlamaForCausalLM, GenerationConfig tokenizer = LlamaTokenizer.from_pretrained("decapoda-research/llama-7b-hf") BASE_MODEL = "decapoda-research/llama-7b-hf" LORA_WEIGHTS = "tloen/alpaca-lora-7b" if torch.cuda.is_available(): device = "cuda" else: device = "cpu" try: if torch.backends.mps.is_available(): device = "mps" except: pass if device == "cuda": model = LlamaForCausalLM.from_pretrained( BASE_MODEL, load_in_8bit=False, torch_dtype=torch.float16, device_map="auto", ) model = PeftModel.from_pretrained( model, LORA_WEIGHTS, torch_dtype=torch.float16, force_download=True ) elif device == "mps": model = LlamaForCausalLM.from_pretrained( BASE_MODEL, device_map={"": device}, torch_dtype=torch.float16, ) model = PeftModel.from_pretrained( model, LORA_WEIGHTS, device_map={"": device}, torch_dtype=torch.float16, ) else: model = LlamaForCausalLM.from_pretrained( BASE_MODEL, device_map={"": device}, low_cpu_mem_usage=True ) model = PeftModel.from_pretrained( model, LORA_WEIGHTS, device_map={"": device}, ) if device != "cpu": model.half() model.eval() if torch.__version__ >= "2": model = torch.compile(model) def evaluate( table, question, input=None, temperature=0.1, top_p=0.75, top_k=40, num_beams=4, max_new_tokens=128, **kwargs, ): prompt = _TEMPLATE + "\n" + _add_markup(table) + "\n" + "Q: " + question + "\n" + "A:" inputs = tokenizer(prompt, return_tensors="pt") input_ids = inputs["input_ids"].to(device) generation_config = GenerationConfig( temperature=temperature, top_p=top_p, top_k=top_k, num_beams=num_beams, **kwargs, ) with torch.no_grad(): generation_output = model.generate( input_ids=input_ids, generation_config=generation_config, return_dict_in_generate=True, output_scores=True, max_new_tokens=max_new_tokens, ) s = generation_output.sequences[0] output = tokenizer.decode(s) #return output.split("A:")[-1].strip() return output ## deplot models model_deplot = Pix2StructForConditionalGeneration.from_pretrained("google/deplot", torch_dtype=torch.bfloat16).to(0) processor_deplot = Pix2StructProcessor.from_pretrained("google/deplot") def process_document(image, question): # image = Image.open(image) inputs = processor_deplot(images=image, text="Generate the underlying data table for the figure below:", return_tensors="pt").to(torch.bfloat16, 0) predictions = model_deplot.generate(**inputs, max_new_tokens=512) table = processor_deplot.decode(predictions[0], skip_special_tokens=True).replace("<0x0A>", "\n") # send prompt+table to LLM res = evaluate(table, question) #return res + "\n\n" + res.split("A:")[-1] return [table, res.split("A:")[-1]] description = "Demo for DePlot+LLM for QA and summarisation. [DePlot](https://arxiv.org/abs/2212.10505) is an image-to-text model that converts plots and charts into a textual sequence. The sequence then is used to prompt LLM for chain-of-thought reasoning. The current underlying LLM is [alpaca-lora](https://huggingface.co/spaces/tloen/alpaca-lora). To use it, simply upload your image and type a question or instruction and click 'submit', or click one of the examples to load them. Read more at the links below." article = "

DePlot: One-shot visual language reasoning by plot-to-table translation

" demo = gr.Interface( fn=process_document, inputs=["image", "text"], outputs=[ gr.inputs.Textbox( lines=8, label="Intermediate Table", ), gr.inputs.Textbox( lines=5, label="Output", ) ], title="DePlot+LLM (Multimodal chain-of-thought reasoning on plots)", description=description, article=article, enable_queue=True, examples=[["deplot_case_study_m1.png", "What is the sum of numbers of Indonesia and Ireland? Remember to think step by step."], ["deplot_case_study_m1.png", "Summarise the chart for me please."], ["deplot_case_study_3.png", "By how much did China's growth rate drop? Think step by step."], ["deplot_case_study_4.png", "How many papers are submitted in 2020?"], ["deplot_case_study_x2.png", "Summarise the chart for me please."]], cache_examples=True) demo.launch(debug=True)