File size: 6,889 Bytes
aa1c44a |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 |
import streamlit as st
import os
import gc
import base64
import tempfile
import uuid
from llama_index import VectorStoreIndex, SimpleDirectoryReader, ServiceContext
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.llms.huggingface import HuggingFaceLLM
from llama_index.prompts import PromptTemplate
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
# ----------------------------
# 1) LLM LOADING
# ----------------------------
@st.cache_resource
def load_llm():
"""
Load the DeepSeek-R1 700B (approx) model from Hugging Face,
using 4-bit quantization and auto device mapping.
"""
model_id = "deepseek-ai/DeepSeek-R1"
# tokenizer
tokenizer = AutoTokenizer.from_pretrained(
model_id,
trust_remote_code=True
)
# model in 4-bit
model = AutoModelForCausalLM.from_pretrained(
model_id,
trust_remote_code=True,
device_map="auto", # auto-shard across all available GPUs
load_in_4bit=True, # bitsandbytes 4-bit quantization
torch_dtype=torch.float16
)
# wrap with LlamaIndex's HuggingFaceLLM
llm = HuggingFaceLLM(
model=model,
tokenizer=tokenizer,
streaming=True,
temperature=0.7,
max_new_tokens=512
)
return llm
# ----------------------------
# 2) STREAMLIT + INDEX SETUP
# ----------------------------
if "id" not in st.session_state:
st.session_state.id = uuid.uuid4()
st.session_state.file_cache = {}
def reset_chat():
st.session_state.messages = []
gc.collect()
def display_pdf(file):
st.markdown("### PDF Preview")
base64_pdf = base64.b64encode(file.read()).decode("utf-8")
pdf_display = f"""
<iframe src="data:application/pdf;base64,{base64_pdf}"
width="400" height="100%"
style="height:100vh; width:100%">
</iframe>
"""
st.markdown(pdf_display, unsafe_allow_html=True)
# Sidebar for file upload
with st.sidebar:
st.header("Add your documents!")
uploaded_file = st.file_uploader("Choose a `.pdf` file", type="pdf")
if uploaded_file:
try:
# Indexing the doc
with tempfile.TemporaryDirectory() as temp_dir:
file_path = os.path.join(temp_dir, uploaded_file.name)
with open(file_path, "wb") as f:
f.write(uploaded_file.getvalue())
file_key = f"{st.session_state.id}-{uploaded_file.name}"
st.write("Indexing your document...")
if file_key not in st.session_state.get('file_cache', {}):
if os.path.exists(temp_dir):
loader = SimpleDirectoryReader(
input_dir=temp_dir,
required_exts=[".pdf"],
recursive=True
)
else:
st.error("Could not find the file. Please reupload.")
st.stop()
docs = loader.load_data()
# Load the HF-based LLM (DeepSeek-R1)
llm = load_llm()
# HuggingFace Embeddings for the VectorStore
embed_model = HuggingFaceEmbedding(
model_name="answerdotai/ModernBERT-large",
trust_remote_code=True
)
# create a service context
service_context = ServiceContext.from_defaults(
llm=llm,
embed_model=embed_model
)
# build the index
index = VectorStoreIndex.from_documents(
docs,
service_context=service_context,
show_progress=True
)
query_engine = index.as_query_engine(streaming=True)
# custom QA prompt
qa_prompt_tmpl_str = (
"Context information is below.\n"
"---------------------\n"
"{context_str}\n"
"---------------------\n"
"Given the context info above, provide a concise answer.\n"
"If you don't know, say 'I don't know'.\n"
"Query: {query_str}\n"
"Answer: "
)
qa_prompt = PromptTemplate(qa_prompt_tmpl_str)
query_engine.update_prompts(
{"response_synthesizer:text_qa_template": qa_prompt}
)
# store in session state
st.session_state.file_cache[file_key] = query_engine
else:
query_engine = st.session_state.file_cache[file_key]
st.success("Ready to Chat!")
display_pdf(uploaded_file)
except Exception as e:
st.error(f"An error occurred: {e}")
st.stop()
col1, col2 = st.columns([6, 1])
with col1:
st.markdown("# RAG with DeepSeek-R1 (700B)")
with col2:
st.button("Clear ↺", on_click=reset_chat)
# Initialize chat if needed
if "messages" not in st.session_state:
reset_chat()
# Render past messages
for message in st.session_state.messages:
with st.chat_message(message["role"]):
st.markdown(message["content"])
# Chat input
if prompt := st.chat_input("Ask a question about your PDF..."):
st.session_state.messages.append({"role": "user", "content": prompt})
with st.chat_message("user"):
st.markdown(prompt)
# Retrieve the engine
if uploaded_file:
file_key = f"{st.session_state.id}-{uploaded_file.name}"
query_engine = st.session_state.file_cache.get(file_key)
else:
query_engine = None
# If no docs, just return a quick message
if not query_engine:
answer = "No documents indexed. Please upload a PDF first."
st.session_state.messages.append({"role": "assistant", "content": answer})
with st.chat_message("assistant"):
st.markdown(answer)
else:
with st.chat_message("assistant"):
message_placeholder = st.empty()
full_response = ""
# Streaming generator from LlamaIndex
streaming_response = query_engine.query(prompt)
for chunk in streaming_response.response_gen:
full_response += chunk
message_placeholder.markdown(full_response + "▌")
message_placeholder.markdown(full_response)
st.session_state.messages.append({"role": "assistant", "content": full_response})
|