tensorwich 0.1.1

A library for saving embeddings for pandas dataframes

tensorwich

Tensorwich is a python package for saving and loading embedding data for pandas dataframe columns as a compact binary format

Features

• Save and load embedding data for pandas dataframe columns as a compact binary format that accompanies the saved csv file

Installation

pip install tensorwich

Usage

import tensorwich as tw
import pandas as pd

# load a df from a csv file and wrap it in a tensorwich
df = pd.read_csv("examples/example_df.csv")
tdf = tw.tensorwich(df)

# add the embedding to the tensorwich/df by using a custom function that generates embeddings
# in this case, we are using the "get_embeddings" function to generate embeddings from the Description column
# of the df
tdf.add_embedding('short_desc_embedding', get_embeddings(tdf.df['Description'].tolist()))

# save the df and related tensors to disk
tdf.save("examples/example_df_with_embeddings.csv")

# load the related tensors from disk
result = tw.tensorwich.load("examples/example_df_with_embeddings.csv")

License

This project is licensed under the MIT License - see the LICENSE file for details