anishde's picture
Update app.py
43d73b0 verified
raw
history blame
12.9 kB
from langchain.document_loaders import PyPDFLoader, TextLoader, Docx2txtLoader
from langchain.text_splitter import CharacterTextSplitter
from langchain.document_loaders import PDFMinerLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import HuggingFaceEmbeddings
from langchain import HuggingFaceHub
from langchain.chains.summarize import load_summarize_chain
from langchain.chains.llm_summarization_checker.base import LLMSummarizationCheckerChain
from langchain.prompts import PromptTemplate
import os
import gradio as gr
import shutil
import re
import tempfile
from pathlib import Path
api_token=os.environ['api']
os.environ["HUGGINFACEHUB_API_TOKEN"]=api_token
# api=userdata.get('api')
# api_token=api
# # api_token =
# os.environ["HUGGINFACEHUB_API_TOKEN"]=api_token
temp_dir = "/content/sample_data"
# file_path_dummy = "/content/2401.10231.pdf"
# if file_path_dummy.lower().endswith(".pdf") :
# loader = TextLoader(file_path_dummy)
# document= loader.load()
# print(document)
def data_ingestion(file_path):
if not os.path.exists(file_path):
raise ValueError(f"File path {file_path} does not exist.")
path = Path(file_path)
file_ext = path.suffix
# file_ext = os.path.splitext(file_path)[-1]
# if file_ext == ".pdf":
if file_path.lower().endswith(".pdf"):
loader = PDFMinerLoader(file_path)
elif file_path.lower().endswith(".txt"):
loader = TextLoader(file_path)
else:
loader = Docx2txtLoader(file_path)
# document= loader.load()
# loader = PDFMinerLoader(file_path)
document= loader.load()
length = len(document[0].page_content)
# Replace CharacterTextSplitter with RecursiveCharacterTextSplitter
text_splitter = RecursiveCharacterTextSplitter(chunk_size=0.03*length, chunk_overlap=0)
split_docs = text_splitter.split_documents(document)
embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2", model_kwargs={'device': 'cpu'})
llm = HuggingFaceHub(repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1",
model_kwargs={"temperature":1, "max_length":10000},
huggingfacehub_api_token=api_token)
return split_docs
# text_splitter = CharacterTextSplitter.from_tiktoken_encoder(
# chunk_size=2000, chunk_overlap=0
# )
# split_docs = text_splitter.split_documents(document)
# documents=split_text_into_batches(str(document),400)
# len(documents)
# documents[0]
# #
# from langchain.text_splitter import CharacterTextSplitter
# text_splitter = CharacterTextSplitter(chunk_size=200, chunk_overlap=0)
# documents = text_splitter.split_documents(document)
# Embeddings
# from langchain.chains.question_answering import load_qa_chain
########## CHAIN 1 norm text
def chain1():
prompt_template = """Please provide a summary of the given study material. Summarize the key concepts, findings, and important details.
Pay special attention to any definitions, theories, or conclusions presented in the text.
Your summary should be concise yet comprehensive, capturing the main points of the study material.
Your job is to write a summary of the document such that every summary of the text is of 2 sentences.
here is the content of the section:
"{text}"
SUMMARY:"""
prompt = PromptTemplate.from_template(prompt_template)
refine_template = (
"Your job is to produce a final summary\n"
# "We have provided an existing summary up to a certain point: {existing_answer}\n"
"We have the opportunity to refine the existing summary"
"(only if needed) with some more context below.\n"
"------------\n"
"{text}\n"
"------------\n"
"Given the new context, refine the original summary in English"
"If the context isn't useful, return the original summary." )
refine_prompt = PromptTemplate.from_template(refine_template)
chain1 = load_summarize_chain(
llm=HuggingFaceHub(repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1",
model_kwargs={"temperature":1, "max_length":10000},
huggingfacehub_api_token=api_token),
chain_type="refine",
question_prompt=prompt,
# refine_prompt=refine_prompt,
return_intermediate_steps=False,
input_key="input_documents",
output_key="output_text",
)
return chain1
# result = chain({"input_documents":split_docs}, return_only_outputs=True)
########## CHAIN 2 Legal Document
def chain2():
prompt_template = """Summarize the provided legal document. Highlight key legal arguments, decisions, and any crucial precedents mentioned.
Include a concise overview of the case or legal matter, focusing on the most significant details.
Ensure the summary captures the essential legal aspects and implications of the document.
Your job is to write a summary of the document such that every summary of the text is of 2 sentences.
here is the content of the section:
"{text}"
SUMMARY:"""
prompt = PromptTemplate.from_template(prompt_template)
refine_template = (
"Your job is to produce a final summary\n"
# "We have provided an existing summary up to a certain point: {existing_answer}\n"
"We have the opportunity to refine the existing summary"
"(only if needed) with some more context below.\n"
"------------\n"
"{text}\n"
"------------\n"
"Given the new context, refine the original summary in English"
"If the context isn't useful, return the original summary." )
refine_prompt = PromptTemplate.from_template(refine_template)
chain2 = load_summarize_chain(
llm = HuggingFaceHub(repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1",
model_kwargs={"temperature":1, "max_length":10000},
huggingfacehub_api_token=api_token),
chain_type = "refine",
question_prompt = prompt,
# refine_prompt = refine_prompt,
return_intermediate_steps=False,
input_key="input_documents",
output_key="output_text",
)
return chain2
# result = chain({"input_documents":split_docs}, return_only_outputs=True)
########## CHAIN 3 arxiv_paper_1
def chain3():
prompt_template = """You are being given a markdown document with headers, this is part of a larger arxiv paper.
Provide a summary of the given research paper. Capture the main research question, methodology, key findings, and any novel contributions made by the paper.
Emphasize the significance of the research and its potential implications.
The summary should be succinct while conveying the essential information presented in the research paper.
Your job is to write a summary of the document such that every summary of the text is of 2 sentences.
here is the content of the section:
"{text}"
SUMMARY:"""
prompt = PromptTemplate.from_template(prompt_template)
refine_template = ("""You are presented with a collection of text snippets. Each snippet is a summary of a specific section from an academic paper published on arXiv. Your objective is to synthesize these snippets into a coherent, concise summary of the entire paper.
DOCUMENT SNIPPETS:
"{text}"
INSTRUCTIONS: Craft a concise summary below, capturing the essence of the paper based on the provided snippets.
It is also important that you highlight the key contributions of the paper, and 3 key takeaways from the paper.
Lastly you should provide a list of 5 questions that you would ask the author of the paper if you had the chance. Remove all the backslash n (\n)
SUMMARY:
"""
)
refine_prompt = PromptTemplate.from_template(refine_template)
chain3 = load_summarize_chain(
llm=HuggingFaceHub(repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1",
model_kwargs={"temperature":1, "max_length":10000},
huggingfacehub_api_token=api_token),
chain_type="refine",
question_prompt=prompt,
# refine_prompt=refine_prompt,
return_intermediate_steps=False,
input_key="input_documents",
output_key="output_text",
)
return chain3
# result = chain({"input_documents":split_docs}, return_only_outputs=True)
# chain.run(document)
# print(result["output_text"])
def chain_function(checkbox_values):
if "Research Paper" in checkbox_values:
output = chain3()
elif "Legal Document" in checkbox_values:
output = chain2()
elif "General Text" in checkbox_values:
output = chain1()
else:
output = "Please select a document type to run."
return output
def result(chain, split_docs):
summaries = []
for doc in split_docs:
result = chain({"input_documents": [doc]})
# result = chain({"input_documents": [doc]}, return_only_outputs=True)
summaries.append(result["output_text"])
text_concat = ""
for i in summaries:
text_concat += i
# output = re.sub(r'\n'," "," ",text_concat)
return text_concat
title = """<p style="font-family:Century Gothic; text-align:center; font-size: 100px">S I M P L I F Y</p>"""
# description = r"""<p style="font-family: Century Gothic; text-align:center; font-size: 100px">S I M P L I F Y</p>
# """
# article = r"""
# If PhotoMaker is helpful, please help to ⭐ the <a href='https://github.com/TencentARC/PhotoMaker' target='_blank'>Github Repo</a>. Thanks!
# [![GitHub Stars](https://img.shields.io/github/stars/TencentARC/PhotoMaker?style=social)](https://github.com/TencentARC/PhotoMaker)
# ---
# πŸ“ **Citation**
# <br>
# If our work is useful for your research, please consider citing:
# ```bibtex
# @article{li2023photomaker,
# title={PhotoMaker: Customizing Realistic Human Photos via Stacked ID Embedding},
# author={Li, Zhen and Cao, Mingdeng and Wang, Xintao and Qi, Zhongang and Cheng, Ming-Ming and Shan, Ying},
# booktitle={arXiv preprint arxiv:2312.04461},
# year={2023}
# }
# ```
# πŸ“‹ **License**
# <br>
# Apache-2.0 LICENSE. Please refer to the [LICENSE file](https://huggingface.co/TencentARC/PhotoMaker/blob/main/LICENSE) for details.
# πŸ“§ **Contact**
# <br>
# If you have any questions, please feel free to reach me out at <b>zhenli1031@gmail.com</b>.
# """
# tips = r"""
# ### Usage tips of PhotoMaker
# 1. Upload more photos of the person to be customized to **improve ID fidelty**. If the input is Asian face(s), maybe consider adding 'asian' before the class word, e.g., `asian woman img`
# 2. When stylizing, does the generated face look too realistic? Adjust the **Style strength** to 30-50, the larger the number, the less ID fidelty, but the stylization ability will be better.
# 3. If you want to generate realistic photos, you could try switching to our other gradio application [PhotoMaker](https://huggingface.co/spaces/TencentARC/PhotoMaker).
# 4. For **faster** speed, reduce the number of generated images and sampling steps. However, please note that reducing the sampling steps may compromise the ID fidelity.
# """
# def process_file(file_obj):
# destination_path = "/content/sample_data" # Replace with your desired path
# shutil.copy(file_obj, destination_path) # Save file to specified path
# return os.path.join(destination_path, file_obj)
def process_file(list_file_obj):
# list_file_path = [x.name for x in list_file_obj if x is not None]
# file_content = file_obj.data
# with tempfile.TemporaryFile() as temp_file:
# temp_file.write(file_content)
# temp_file_path = temp_file.name
return list_file_obj[0].name
def inference(checkbox_values, uploaded_file):
file_path = process_file(uploaded_file)
split_docs = data_ingestion(file_path)
chain = chain_function(checkbox_values)
summary = result(chain, split_docs)
return summary
with gr.Blocks(theme="monochrome") as demo:
gr.Markdown(title)
with gr.Row():
with gr.Column():
checkbox_values = gr.CheckboxGroup(["Research Paper", "Legal Document", "General Text"], label="Choose the document type")
uploaded_file = gr.Files(height=100, file_count="multiple", file_types=["text", ".docx", "pdf"], interactive=True, label="Upload your File.")
btn = gr.Button("Submit") # Place the button outside the Row for vertical alignment
with gr.Column():
txt = gr.Textbox(
show_label=False,
# placeholder="Simplify."
)
btn.click(
fn=inference,
inputs=[checkbox_values, uploaded_file],
outputs=[txt],
queue=False
)
if __name__ == "__main__":
demo.queue()
demo.launch(debug = True)