File size: 2,059 Bytes
aabdf81 8314602 5bbc60d aabdf81 5bbc60d 8314602 |
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 |
import shutil
from haystack.document_stores import FAISSDocumentStore
from haystack.nodes import EmbeddingRetriever
from haystack.pipelines import ExtractiveQAPipeline
from haystack.nodes import FARMReader
import streamlit as st
from config import (INDEX_DIR, RETRIEVER_MODEL, RETRIEVER_MODEL_FORMAT,
READER_MODEL, READER_CONFIG_THRESHOLD, QUESTIONS_PATH)
# cached to make index and models load only at start
@st.cache(hash_funcs={"builtins.SwigPyObject": lambda _: None},
allow_output_mutation=True)
def start_haystack():
"""
load document store, retriever, reader and create pipeline
"""
shutil.copy(f'{INDEX_DIR}/faiss_document_store.db', '.')
document_store = FAISSDocumentStore(
faiss_index_path=f'{INDEX_DIR}/my_faiss_index.faiss',
faiss_config_path=f'{INDEX_DIR}/my_faiss_index.json')
print(f'Index size: {document_store.get_document_count()}')
retriever = EmbeddingRetriever(
document_store=document_store,
embedding_model=RETRIEVER_MODEL,
model_format=RETRIEVER_MODEL_FORMAT
)
reader = FARMReader(model_name_or_path=READER_MODEL,
use_gpu=False,
confidence_threshold=READER_CONFIG_THRESHOLD)
pipe = ExtractiveQAPipeline(reader, retriever)
return pipe
pipe = start_haystack()
# the pipeline is not included as parameter of the following function,
# because it is difficult to cache
@st.cache(persist=True, allow_output_mutation=True)
def query(question: str, retriever_top_k: int = 10, reader_top_k: int = 5):
"""Run query and get answers"""
params = {"Retriever": {"top_k": retriever_top_k},
"Reader": {"top_k": reader_top_k}}
results = pipe.run(question, params=params)
return results
@st.cache()
def load_questions():
"""Load selected questions from file"""
with open(QUESTIONS_PATH) as fin:
questions = [line.strip() for line in fin.readlines()
if not line.startswith('#')]
return questions
|