File size: 1,746 Bytes
d3a1fe2 |
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 |
from langchain_community.document_loaders import DirectoryLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_chroma import Chroma
import os
import shutil
def create_db(
chunk_size,
chunk_overlap,
INPUT_PATH="./data/books/",
INPUT_GLOB=["*.txt", "*.md"],
MODEL_NAME="Alibaba-NLP/gte-multilingual-base",
CHROMA_PATH="./chromadb/",
):
# setup embeddings
embeddings = HuggingFaceEmbeddings(
model_name=MODEL_NAME,
model_kwargs={"device": "cuda", "trust_remote_code": True},
encode_kwargs={"normalize_embeddings": True},
)
# load documents
raw_documents = DirectoryLoader(INPUT_PATH, glob=INPUT_GLOB).load()
text_splitter = RecursiveCharacterTextSplitter(
chunk_size=chunk_size,
chunk_overlap=chunk_overlap,
length_function=len,
add_start_index=True,
)
documents = text_splitter.split_documents(raw_documents)
print(f"Split {len(raw_documents)} documents into {len(documents)} chunks.")
# Clear out the database first.
if os.path.exists(CHROMA_PATH):
shutil.rmtree(CHROMA_PATH)
# Create a new DB from the documents.
db = Chroma.from_documents(
documents,
embeddings,
persist_directory=CHROMA_PATH,
collection_metadata={"hnsw:space": "cosine"},
)
print(f"Saved {len(documents)} chunks to {CHROMA_PATH}.")
return db
if __name__ == "__main__":
create_db(
1000,
500,
INPUT_PATH="./data/books/dracula_segmented/",
INPUT_GLOB=["*.txt"],
MODEL_NAME="Alibaba-NLP/gte-multilingual-base",
CHROMA_PATH="./chromadb/",
)
|