I’m confused as to what the encoder_no_repeat_ngram_size
argument does. The documentation describes it like so:
Value that will be used by default in the
generate
method of the model forencoder_no_repeat_ngram_size
. If set to int > 0, all ngrams of that size that occur in theencoder_input_ids
cannot occur in thedecoder_input_ids
.
I understand that decoder_input_ids
are labels for training the decoder, what’s their purpose in generation?