The Decoder is one of the two main blocks of the original Transformer Architecture. It takes the Vector Embeddings generated by the encoder, along with the partial output text generated so far, to predict the next word in the sequence (translation or text generation).
Context:
- The main purpose of the decoder is to generate the output text from the embedding vectors and from the partial output which it has received.
- The GPT architecture consists only of the decoder block without the encoder block.
