Struct ProphetNetGenerationOutput

pub struct ProphetNetGenerationOutput {
    pub logits: Tensor,
    pub ngram_logits: Option<Tensor>,
    pub ngram_hidden_states: Option<Tensor>,
    pub all_decoder_hidden_states: Option<Vec<Tensor>>,
    pub all_ngram_hidden_states: Option<Vec<Tensor>>,
    pub all_attentions: Option<Vec<Tensor>>,
    pub all_ngram_attentions: Option<Vec<Tensor>>,
    pub all_cross_attentions: Option<Vec<Tensor>>,
    pub next_decoder_cache: Option<Vec<(Option<LayerState>, Option<LayerState>)>>,
}

Container holding a ProphetNet model generation output

Fields

logits: Tensor

Prediction logits

ngram_logits: Option<Tensor>

Ngram prediction logits

ngram_hidden_states: Option<Tensor>

last decoder layer ngram hidden state

all_decoder_hidden_states: Option<Vec<Tensor>>

Hidden states for all intermediate layers

all_ngram_hidden_states: Option<Vec<Tensor>>

Hidden states (ngram) for all intermediate layers

all_attentions: Option<Vec<Tensor>>

Attention weights for all intermediate layers

all_ngram_attentions: Option<Vec<Tensor>>

Ngram attention weights for all intermediate layers

all_cross_attentions: Option<Vec<Tensor>>

Cross attention weights for all intermediate layers

next_decoder_cache: Option<Vec<(Option<LayerState>, Option<LayerState>)>>

Cached outputs of the model (attention layers keys and values) if the model is used for generation