SearchingFace / vectorize_dataset.py
nkasmanoff's picture
Update vectorize_dataset.py
8ac2414
raw
history blame contribute delete
No virus
1.47 kB
from datasets import load_dataset
from helpers import clean_up_tags
from langchain.text_splitter import CharacterTextSplitter
from langchain.vectorstores import Chroma
from langchain.document_loaders import DataFrameLoader
def load_descriptions_data(dataset='nkasmanoff/hf-dataset-cards'):
if dataset == 'hf-dataset-cards':
hf_datasets = load_dataset(dataset)
hf_df = hf_datasets['train'].to_pandas()
hf_df.dropna(subset=['README'],inplace=True)
hf_df['description_full'] = hf_df['README']
else:
hf_datasets = load_dataset('nkasmanoff/huggingface-datasets')
hf_df = hf_datasets['train'].to_pandas()
hf_df['tags_cleaned'] = hf_df['tags'].apply(clean_up_tags)
hf_df.dropna(subset=['description'],inplace=True)
hf_df['description_full'] = hf_df['description'].fillna('') + ' ' + hf_df['tags_cleaned']
hf_df = hf_df[hf_df['description_full'] != ' ']
hf_df = hf_df[['id','description_full']]
return hf_df
def create_db(hf_df, embeddings):
loader = DataFrameLoader(hf_df, page_content_column="description_full")
documents = loader.load()
# split the documents into chunks
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
texts = text_splitter.split_documents(documents)
# select which embeddings we want to use
# create the vectorestore to use as the index
db = Chroma.from_documents(texts, embeddings)
return db