Spaces:
Sleeping
Sleeping
File size: 3,520 Bytes
32a6937 01f4bd7 32a6937 01f4bd7 32a6937 |
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 |
"""Main entrypoint for the app."""
import os
from timeit import default_timer as timer
from typing import List, Optional
from dotenv import find_dotenv, load_dotenv
from langchain_community.embeddings import HuggingFaceInstructEmbeddings
from langchain.vectorstores.chroma import Chroma
from langchain.vectorstores.faiss import FAISS
from app_modules.llm_loader import LLMLoader
from app_modules.utils import get_device_types, init_settings
found_dotenv = find_dotenv(".env")
if len(found_dotenv) == 0:
found_dotenv = find_dotenv(".env.example")
print(f"loading env vars from: {found_dotenv}")
load_dotenv(found_dotenv, override=False)
# Constants
init_settings()
if os.environ.get("LANGCHAIN_DEBUG") == "true":
from langchain.globals import set_debug
set_debug(True)
from app_modules.llm_qa_chain import QAChain
from app_modules.llm_chat_chain import ChatChain
import nltk
def app_init():
# https://github.com/huggingface/transformers/issues/17611
os.environ["CURL_CA_BUNDLE"] = ""
nltk.download("punkt")
hf_embeddings_device_type, hf_pipeline_device_type = get_device_types()
print(f"hf_embeddings_device_type: {hf_embeddings_device_type}")
print(f"hf_pipeline_device_type: {hf_pipeline_device_type}")
hf_embeddings_model_name = (
os.environ.get("HF_EMBEDDINGS_MODEL_NAME") or "hkunlp/instructor-xl"
)
n_threds = int(os.environ.get("NUMBER_OF_CPU_CORES") or "4")
index_path = os.environ.get("FAISS_INDEX_PATH") or os.environ.get(
"CHROMADB_INDEX_PATH"
)
using_faiss = os.environ.get("FAISS_INDEX_PATH") is not None
llm_model_type = os.environ.get("LLM_MODEL_TYPE")
qa_with_rag = os.getenv("QA_WITH_RAG", "true").lower() == "true"
print(f"qa_with_rag: {qa_with_rag}")
retrieve_from_questions_file = os.getenv("RETRIEVER_TYPE") == "questions_file"
print(f"retrieve_from_questions_file: {retrieve_from_questions_file}", flush=True)
if qa_with_rag and not retrieve_from_questions_file:
print(f"hf_embeddings_model_name: {hf_embeddings_model_name}")
start = timer()
embeddings = HuggingFaceInstructEmbeddings(
model_name=hf_embeddings_model_name,
model_kwargs={"device": hf_embeddings_device_type},
)
end = timer()
print(f"Completed in {end - start:.3f}s")
vectorstore = None
if qa_with_rag and not retrieve_from_questions_file:
start = timer()
print(
f"Load index from {index_path} with {'FAISS' if using_faiss else 'Chroma'}"
)
if not os.path.isdir(index_path):
raise ValueError(f"{index_path} does not exist!")
elif using_faiss:
vectorstore = FAISS.load_local(
index_path, embeddings, allow_dangerous_deserialization=True
)
else:
vectorstore = Chroma(
embedding_function=embeddings, persist_directory=index_path
)
end = timer()
print(f"Completed in {end - start:.3f}s")
start = timer()
llm_loader = LLMLoader(llm_model_type)
llm_loader.init(n_threds=n_threds, hf_pipeline_device_type=hf_pipeline_device_type)
if qa_with_rag:
qa_chain = QAChain(vectorstore, llm_loader)
else:
qa_chain = ChatChain(llm_loader)
end = timer()
print(f"Completed in {end - start:.3f}s")
return llm_loader, qa_chain
|