Learn how seq2seq data is prepared: the dialogue (with the prefix 'summarize: ' for T5) becomes the input, the summary becomes the labels; both are tokenised and cut to their max length. A seq2seq collator pads inputs and labels in each batch.
Notes, key code, what to skip and 2 curated resources are inside the platform.Join the waitlist and we'll build you a roadmap that includes this topic only if your goal needs it.