infer function

cn3.rs

    pub fn infer_sentence_embedding(&self, sentence: &str) 
      -> anyhow::Result<Tensor> {
        let tokens = self
            .tokenizer
            .encode(sentence, true)
            .map_err(anyhow::Error::msg)?;
        let token_ids = Tensor::new(tokens.get_ids(), &self.device)?.unsqueeze(0)?;
        let token_type_ids = token_ids.zeros_like()?;
        let start = std::time::Instant::now();
        let embeddings = self.model.forward(&token_ids, &token_type_ids)?;
        println!("time taken for forward: {:?}", start.elapsed());
        println!("embeddings: {:?}", embeddings);
        let embeddings = Self::apply_max_pooling(&embeddings)?;
        println!("embeddings after pooling: {:?}", embeddings);
        let embeddings = Self::l2_normalize(&embeddings)?;
        Ok(embeddings)
    }