Differences

This shows you the differences between two versions of the page.

--- nlp:non-autoregressive_seq2seq [2021/07/13 20:17] – [Autoregressive vs Non-Autoregressive] jmflanig
+++ nlp:non-autoregressive_seq2seq [2024/05/03 03:37] (current) – [Key Papers] jmflanig
@@ Line 6: / Line 6: @@
 generates tokens conditioned on the sequence of tokens previously generated.  In other words, it operates one step at a time: it generates each token conditioned on the sequence of tokens previously generated. Examples of autoregressive models include RNNs, LSTMs, CNNs (masked convolution layers), and Transformers (from [[https://arxiv.org/pdf/1711.02281.pdf|Gu 2017]]).
-**Definition: Non-Autoregressive ([[https://arxiv.org/pdf/1711.02281.pdf|Gu 2017]]):** A non-autoregressive model removes the conditional dependence between output tokens and generates them in parallel.  For an example see equation 3 of [[https://arxiv.org/pdf/1711.02281.pdf|Gu 2017]].
+**Definition: Non-Autoregressive ([[https://arxiv.org/pdf/1711.02281.pdf|Gu 2017]]):** A non-autoregressive model removes the conditional dependence between output tokens and generates them in parallel.  See equation 3 of [[https://arxiv.org/pdf/1711.02281.pdf|Gu 2017]] for an example.
-**Note:** There are also **global models** like [[ml:Conditional Random Field|Conditional Random Fields]], which are not autoregressive and not non-autoregressive by the above definitions.  Instead they perform inference (dynamic programming, MCMC, etc) to maximize a global scoring function.
+**Note:** There are also **global models** like [[ml:Conditional Random Field|Conditional Random Fields]], which are not autoregressive and not non-autoregressive by the above definitions.  Instead they perform inference (using dynamic programming, MCMC, etc) to maximize a global scoring function.
 ===== Summary =====
@@ Line 36: / Line 36: @@
   * [[https://arxiv.org/pdf/1711.02281.pdf|Gu et al 2017 - Non-Autoregressive Neural Machine Translation]]
+  * [[http://proceedings.mlr.press/v80/kaiser18a/kaiser18a.pdf|Kaiser et al 2018 - Fast Decoding in Sequence Models Using Discrete Latent Variables]]
   * **[[https://arxiv.org/pdf/1904.09324.pdf|Ghazvininejad et al 2019 - Mask-Predict: Parallel Decoding of Conditional Masked Language Models]]**
   * [[https://arxiv.org/pdf/2002.07233.pdf|Lee et al 2020 - On the Discrepancy between Density Estimation and Sequence Generation]]
@@ Line 44: / Line 45: @@
   * **[[https://arxiv.org/pdf/2004.07437.pdf|Non-Autoregressive Machine Translation with Latent Alignments]]** From Google, uses CTC loss from Gu & Kong 2020
   * [[https://arxiv.org/pdf/2006.10369.pdf|Kasai et al 2020 - Deep Encoder, Shallow Decoder: Reevaluating Non-autoregressive Machine Translation]]
+  * [[https://arxiv.org/pdf/2404.12022|Wu et al 2024 - Parallel Decoding via Hidden Transfer for Lossless Large Language Model Acceleration]]
+===== Papers =====
+  * [[https://arxiv.org/pdf/2305.10427.pdf|Santilli et al 2023 - Accelerating Transformer Inference for Translation via Parallel Decoding]]