File size: 4,304 Bytes
947c232
 
 
 
 
 
 
 
 
 
4dcfe25
947c232
4dcfe25
947c232
4dcfe25
efa6eba
4dcfe25
947c232
 
 
 
efa6eba
947c232
 
4dcfe25
947c232
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4dcfe25
947c232
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4dcfe25
 
 
947c232
 
 
 
 
 
 
 
 
 
 
4dcfe25
947c232
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
import gradio as gr
import asyncio
import json
import os
import pickle
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_core.prompts import PromptTemplate
from langchain_community.document_loaders import PDFMinerLoader, CSVLoader, JSONLoader
from langchain.text_splitter import SentenceTransformersTokenTextSplitter
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, pipeline

MODEL_NAME = "TheBloke/Llama-2-7B-GPTQ"

# Initialize tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, device_map="cpu")

text_pipeline = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer
)

# Define prompt template
template = """
<s>[INST] <<SYS>>
Use the following information to answer the question at the end.
<</SYS>>

{context}

{question} [/INST]
"""
prompt = PromptTemplate(template=template, input_variables=["context", "question"])

# Path to cache files
CACHE_DIR = 'cache'
os.makedirs(CACHE_DIR, exist_ok=True)

def save_cache(filename, data):
    with open(os.path.join(CACHE_DIR, filename), 'wb') as f:
        pickle.dump(data, f)

def load_cache(filename):
    try:
        with open(os.path.join(CACHE_DIR, filename), 'rb') as f:
            return pickle.load(f)
    except FileNotFoundError:
        return None

async def process_files(file_paths):
    try:
        print("Processing files...")
        docs = []
        for file_path in file_paths:
            if file_path.endswith('.pdf'):
                loader = PDFMinerLoader(file_path)
                docs.extend(await asyncio.to_thread(loader.load))
            elif file_path.endswith('.csv'):
                loader = CSVLoader(file_path)
                docs.extend(loader.load())
            elif file_path.endswith('.json'):
                with open(file_path, 'r', encoding='utf-8') as f:
                    data = json.load(f)
                docs.append(data)

        print("Files loaded.")
        text_splitter = SentenceTransformersTokenTextSplitter(chunk_size=1024, chunk_overlap=64)
        texts = text_splitter.split_documents(docs)
        print("Text split into chunks.")

        embeddings = HuggingFaceEmbeddings(
            model_name="thenlper/gte-large",
            model_kwargs={"device": "cpu"},
            encode_kwargs={"normalize_embeddings": True},
        )

        db = FAISS.from_documents(texts, embeddings)
        print("FAISS index created.")
        save_cache('cache_key', (texts, db, embeddings))
        return texts, db, embeddings
    except Exception as e:
        print(f"Error: {e}")
        return None, None, str(e)

async def query_files(files, question):
    if not files or not question.strip():
        return "Please upload valid files and enter a question."

    print("Starting query processing...")
    file_paths = [file.name for file in files]

    texts, db, embeddings = await process_files(file_paths)

    if db is None:
        print("Error during processing.")
        return f"Error processing files: {embeddings}"

    print("Processing complete.")
    results = db.similarity_search(question, k=5)
    context = " ".join([result.page_content for result in results])

    prompt_text = prompt.format(context=context, question=question)

    generated_text = text_pipeline(prompt_text)[0]['generated_text']

    return generated_text

def process_and_query(files, question):
    return asyncio.run(query_files(files, question))

with gr.Blocks() as interface:
    gr.Markdown("### Retrieval Augmented Generation (RAG) for LLM Local Trial")
    gr.Markdown(
        "Upload multiple files (PDF, CSV, JSON) and ask a question. The app will generate the answer based on the content of the input files.")

    with gr.Row():
        question_input = gr.Textbox(label="Enter your question", lines=3)
        files_input = gr.File(label="Upload Files", type="filepath", file_count="multiple")  # Multiple file input

    submit_button = gr.Button("Submit")
    output_text = gr.Textbox(label="LLM Response", lines=8)
    submit_button.click(process_and_query, inputs=[files_input, question_input], outputs=output_text)

if __name__ == "__main__":
    interface.launch()