Trait Gliner2Engine 

pub trait Gliner2Engine: Send + Sync {
    type Tensor: Send;

    // Required methods
    fn dup_tensor(&self, t: &Self::Tensor) -> Self::Tensor;
    fn hidden_size(&self) -> usize;
    fn max_width(&self) -> usize;
    fn encode_sequence(
        &self,
        input_ids: &Self::Tensor,
        attention_mask: &Self::Tensor,
    ) -> Result<Self::Tensor>;
    fn gather_text_word_embeddings(
        &self,
        last_hidden: &Self::Tensor,
        positions: &[usize],
    ) -> Result<Self::Tensor>;
    fn gather_text_word_embeddings_batch_idx(
        &self,
        last_hidden: &Self::Tensor,
        batch_idx: usize,
        positions: &[usize],
    ) -> Result<Self::Tensor>;
    fn compute_span_rep(
        &self,
        text_word_embs: &Self::Tensor,
    ) -> Result<Self::Tensor>;
    fn compute_span_rep_batched(
        &self,
        token_embs_list: &[Self::Tensor],
    ) -> Result<Vec<Self::Tensor>>;
    fn classifier_logits(
        &self,
        label_rows: &Self::Tensor,
    ) -> Result<Self::Tensor>;
    fn count_predict(&self, p_embedding: &Self::Tensor) -> Result<usize>;
    fn span_scores_sigmoid(
        &self,
        span_rep: &Self::Tensor,
        field_embs: &Self::Tensor,
        pred_count: usize,
    ) -> Result<Self::Tensor>;
    fn single_sample_inputs(
        &self,
        input_ids: &[u32],
    ) -> Result<(Self::Tensor, Self::Tensor)>;
    fn batch_inputs(
        &self,
        input_ids: Vec<u32>,
        attention_mask_i64: Vec<i64>,
        batch_size: usize,
        max_seq_len: usize,
    ) -> Result<(Self::Tensor, Self::Tensor)>;
    fn batch_row_hidden(
        &self,
        hidden: &Self::Tensor,
        idx: usize,
    ) -> Result<Self::Tensor>;
    fn stack_schema_token_embeddings(
        &self,
        last_hidden_seq: &Self::Tensor,
        positions: &[usize],
    ) -> Result<Self::Tensor>;
    fn tensor_dim0(&self, t: &Self::Tensor) -> Result<usize>;
    fn tensor_narrow0(
        &self,
        t: &Self::Tensor,
        start: usize,
        len: usize,
    ) -> Result<Self::Tensor>;
    fn tensor_index0(&self, t: &Self::Tensor, i: usize) -> Result<Self::Tensor>;
    fn tensor_logits_1d(&self, logits: &Self::Tensor) -> Result<Vec<f32>>;
    fn tensor_span_scores_to_vec4(
        &self,
        t: &Self::Tensor,
    ) -> Result<Vec<Vec<Vec<Vec<f32>>>>>;
}

Inference backend for the GLiNER2 extract pipeline (encoder + heads + tensor helpers).

Send + Sync is required so that &Engine can be shared across Rayon worker threads for parallel preprocessing and per-record decoding.

Required Associated Types

type Tensor: Send

Required Methods

fn dup_tensor(&self, t: &Self::Tensor) -> Self::Tensor

Cheap copy for backends where Clone is unavailable (e.g. tch::Tensor uses shallow_clone).

fn hidden_size(&self) -> usize

fn max_width(&self) -> usize

fn encode_sequence( &self, input_ids: &Self::Tensor, attention_mask: &Self::Tensor, ) -> Result<Self::Tensor>

fn gather_text_word_embeddings( &self, last_hidden: &Self::Tensor, positions: &[usize], ) -> Result<Self::Tensor>

fn gather_text_word_embeddings_batch_idx( &self, last_hidden: &Self::Tensor, batch_idx: usize, positions: &[usize], ) -> Result<Self::Tensor>

fn compute_span_rep( &self, text_word_embs: &Self::Tensor, ) -> Result<Self::Tensor>

fn compute_span_rep_batched( &self, token_embs_list: &[Self::Tensor], ) -> Result<Vec<Self::Tensor>>

fn classifier_logits(&self, label_rows: &Self::Tensor) -> Result<Self::Tensor>

fn count_predict(&self, p_embedding: &Self::Tensor) -> Result<usize>

Argmax over the count head (20-class) for the [P] embedding row.

fn span_scores_sigmoid( &self, span_rep: &Self::Tensor, field_embs: &Self::Tensor, pred_count: usize, ) -> Result<Self::Tensor>

fn single_sample_inputs( &self, input_ids: &[u32], ) -> Result<(Self::Tensor, Self::Tensor)>

[seq] token ids → [1, seq] int tensors and [1, seq] attention mask (ones).

fn batch_inputs( &self, input_ids: Vec<u32>, attention_mask_i64: Vec<i64>, batch_size: usize, max_seq_len: usize, ) -> Result<(Self::Tensor, Self::Tensor)>

fn batch_row_hidden( &self, hidden: &Self::Tensor, idx: usize, ) -> Result<Self::Tensor>

Row idx of [B, seq, D] encoder output → [seq, D].

fn stack_schema_token_embeddings( &self, last_hidden_seq: &Self::Tensor, positions: &[usize], ) -> Result<Self::Tensor>

Stack last_hidden_seq[pos] for schema special token positions → [n, D].

fn tensor_dim0(&self, t: &Self::Tensor) -> Result<usize>

fn tensor_narrow0( &self, t: &Self::Tensor, start: usize, len: usize, ) -> Result<Self::Tensor>

fn tensor_index0(&self, t: &Self::Tensor, i: usize) -> Result<Self::Tensor>

fn tensor_logits_1d(&self, logits: &Self::Tensor) -> Result<Vec<f32>>

fn tensor_span_scores_to_vec4( &self, t: &Self::Tensor, ) -> Result<Vec<Vec<Vec<Vec<f32>>>>>

Implementors

impl Gliner2Engine for CandleExtractor

type Tensor = Tensor