What is the correct sequence of steps in the process of using a sequence transduction model for translating input sequences into output sequences?

First, the model reads the whole input sentence and turns each word or character into a numerical representation, called embeddings, which it processes with an encoder to capture the meaning of the whole sentence. Next, it starts generating the output word by word, using the encoder’s information and a decoder that predicts each next word based on what has already been produced and the encoded input. During this step the model often uses attention to focus on the most relevant parts of the input while deciding each output word. Finally, it repeats this decoding until it reaches a special end token, giving a complete translated sentence. For example, translating “Hello” to French would involve encoding “Hello,” decoding “Bonjour,” and stopping when the end token appears.