translate.py

import json
from tensorflow.keras.preprocessing.text import tokenizer_from_json
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import load_model
import numpy as np

class TranslationModel:
    def __init__(self, model_path,en_tokenizer,fr_tokenizer,max_len):
        # Load the Keras model
        self.model = load_model(model_path)

        self.english_tokenizer = en_tokenizer

        self.french_tokenizer = fr_tokenizer

        # Define max sequence length
        self.max_seq_length = max_len

    def logits_index(self,text):
        input_sequence = self.english_tokenizer.texts_to_sequences([text])
        padded_input_sequence = pad_sequences(input_sequence, maxlen=self.max_seq_length, padding='post')
        decoded_text = '<start>'

        for i in range(self.max_seq_length):
            target_sequence = self.french_tokenizer.texts_to_sequences([decoded_text])
            padded_target_sequence = pad_sequences(target_sequence, maxlen=self.max_seq_length, padding='post')[:, :-1]

            prediction = self.model([padded_input_sequence, padded_target_sequence])

            idx = np.argmax(prediction[0, i, :])

        return idx