read_dataset.py

import pyarrow.parquet as pq

METADATA_FILEPATH = '/home/sena/plato_backend/scripts/huggingface_embedding_sharing/huggingface_embedding_sharing/save_backup/metadata_file.parquet'
EMBEDDINGS_FILEPATH = '/home/sena/plato_backend/scripts/huggingface_embedding_sharing/huggingface_embedding_sharing/embeddings_2.txt'


def read_embeddings(filepath):

    # Read the lines from the file
    with open(filepath, 'rb') as f:
        lines = f.readlines()
    
    embedding_dict = {}
    
    for index, line in enumerate(lines):
        line_as_list = line.split()
        
        # Decode the byte string to a regular string and convert to float
        float_list = [float(byte_string.decode('utf-8')) for byte_string in line_as_list]
        
        # Save the float list (embeddings) to the dictionary with the index as the key
        embedding_dict[index] = float_list
    
    return embedding_dict


#Read the metadata file
metadata_df = pq.read_table(METADATA_FILEPATH).to_pandas()

## How to reach embeddings - example for a specific row
unique_id = '390U_535R_2021-08-21064631_69'

# Get the row with the unique_id
row = metadata_df.loc[metadata_df['unique_id'] == unique_id]
       
# Get the values for 'data_row' and 'embedding_file'
dat_row_value = row['dat_row'] # index of the row in the embedding file
embedding_file_value = row['embedding_file'] # specific embedding file for the row

# Read the embeddings in the embedding file 
embeddings = read_embeddings(EMBEDDINGS_FILEPATH)
vector = embeddings[int(dat_row_value)]

print(f"Row: {row}, Dat Row: {dat_row_value}, Embedding File: {embedding_file_value}, Vector length: {len(vector)}")
#print(f"Vector: {vector}")