app.py

import gradio as gr
import io
import numpy as np
from tok import Tokenizer

# Vector Loader   
def load_vectors(fname):
    fin = io.open(fname, 'r', encoding='utf-8', newline='\n', errors='ignore')
    data = {}
    for line in fin:
        tokens = line.rstrip().split(' ')
        data[tokens[0]] = np.array(list(map(float, tokens[1:])))  # Convert to NumPy array
    del fin
    return data, sorted(data.keys(), key=len, reverse=True)
vectors, sorted_vector = load_vectors('wiki-news-300d-1M.vec')

# Tokenizer
tokenizer = Tokenizer(protected_words=sorted_vector)
def tokenize(text):
    return tokenizer.word_tokenize(text)

# Interface
def onInput(paragraph, progress = gr.Progress()):
    progress(0, "Tokenizing...")
    tokens = tokenize(paragraph)
    
    progress(0.1, "Initializing merged vector...")
    if not tokens:  # Handle case with no tokens found
        return np.zeros(300).tolist()  # Return a zero vector of appropriate dimension
    
    merged_vector = np.zeros(300)  # Assuming vectors are 300-dimensional
    
    # Merge vectors using NumPy
    totalTokens = len(tokens)
    for ind, token in enumerate(tokens):
        completion = 0.7*((ind+1)/totalTokens)
        progress(0.1 + completion, f"Merging {token}, Token #{tokens.index(token)+1}/{len(tokens)}")

        vector = vectors[token]
        merged_vector += vector
    
    # Normalize
    progress(0.9, "Normalizing...")
    merged_vector /= len(tokens)
    
    progress(1, "Converting to list...")
    return merged_vector.tolist()  # Convert back to list for output

demo = gr.Interface(fn=onInput, inputs="text", outputs="text")
demo.launch()