Spaces:
Sleeping
Sleeping
File size: 5,075 Bytes
9cf8a87 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 |
import gradio as gr
from qdrant_client import models, QdrantClient
from sentence_transformers import SentenceTransformer
from PyPDF2 import PdfReader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.callbacks.manager import CallbackManager
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
# from langchain.llms import LlamaCpp
from langchain.vectorstores import Qdrant
from qdrant_client.http import models
# from langchain.llms import CTransformers
from ctransformers import AutoModelForCausalLM
# loading the embedding model -
encoder = SentenceTransformer('jinaai/jina-embedding-b-en-v1')
print("embedding model loaded.............................")
print("####################################################")
# loading the LLM
callback_manager = CallbackManager([StreamingStdOutCallbackHandler()])
print("loading the LLM......................................")
# llm = LlamaCpp(
# model_path="TheBloke/Llama-2-7B-Chat-GGUF/llama-2-7b-chat.Q8_0.gguf",
# n_ctx=2048,
# f16_kv=True, # MUST set to True, otherwise you will run into problem after a couple of calls
# callback_manager=callback_manager,
# verbose=True,
# )
llm = AutoModelForCausalLM.from_pretrained("TheBloke/Llama-2-7B-Chat-GGUF",
model_file="llama-2-7b-chat.Q3_K_L.gguf",
model_type="llama",
temperature = 0.2,
repetition_penalty = 1.5
)
print("LLM loaded........................................")
print("################################################################")
def get_chunks(text):
text_splitter = RecursiveCharacterTextSplitter(
# seperator = "\n",
chunk_size = 250,
chunk_overlap = 50,
length_function = len,
)
chunks = text_splitter.split_text(text)
return chunks
pdf_path = './100 Weird Facts About the Human Body.pdf'
reader = PdfReader(pdf_path)
text = ""
num_of_pages = len(reader.pages)
for page in range(num_of_pages):
current_page = reader.pages[page]
text += current_page.extract_text()
chunks = get_chunks(text)
print(chunks)
print("Chunks are ready.....................................")
print("######################################################")
client = QdrantClient(path = "./db")
print("db created................................................")
print("#####################################################################")
client.recreate_collection(
collection_name="my_facts",
vectors_config=models.VectorParams(
size=encoder.get_sentence_embedding_dimension(), # Vector size is defined by used model
distance=models.Distance.COSINE,
),
)
print("Collection created........................................")
print("#########################################################")
li = []
for i in range(len(chunks)):
li.append(i)
dic = zip(li, chunks)
dic= dict(dic)
client.upload_records(
collection_name="my_facts",
records=[
models.Record(
id=idx,
vector=encoder.encode(dic[idx]).tolist(),
payload= {dic[idx][:5] : dic[idx]}
) for idx in dic.keys()
],
)
print("Records uploaded........................................")
print("###########################################################")
def chat(question):
hits = client.search(
collection_name="my_facts",
query_vector=encoder.encode(question).tolist(),
limit=3
)
context = []
for hit in hits:
context.append(list(hit.payload.values())[0])
context = context[0] + context[1] + context[2]
system_prompt = """You are a helpful assistant, you will use the provided context to answer user questions.
Read the given context before answering questions and think step by step. If you can not answer a user question based on
the provided context, inform the user. Do not use any other information for answering user. Provide a detailed answer to the question."""
B_INST, E_INST = "[INST]", "[/INST]"
B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"
SYSTEM_PROMPT = B_SYS + system_prompt + E_SYS
instruction = f"""
Context: {context}
User: {question}"""
prompt_template = B_INST + SYSTEM_PROMPT + instruction + E_INST
result = llm(prompt_template)
return result
screen = gr.Interface(
fn = chat,
inputs = gr.Textbox(lines = 10, placeholder = "Enter your question here 👉"),
outputs = gr.Textbox(lines = 10, placeholder = "Your answer will be here soon 🚀"),
title="Q&N with PDF 👩🏻💻📓✍🏻💡",
description="This app facilitates a conversation with PDFs available on https://www.delo.si/assets/media/other/20110728/100%20Weird%20Facts%20About%20the%20Human%20Body.pdf💡",
theme="soft",
# examples=["Hello", "what is the speed of human nerve impulses?"],
)
screen.launch() |