The Universal Transformer is a generalization of the Transformer architecture. Universal Transformers combine the parallelizability and global receptive field of feed-forward sequence models like the Transformer with the recurrent inductive bias of RNNs. They also utilise a dynamic per-position halting mechanism.
Source: Universal TransformersPaper | Code | Results | Date | Stars |
---|
Task | Papers | Share |
---|---|---|
Sentence | 4 | 12.12% |
Language Modelling | 3 | 9.09% |
Text Generation | 2 | 6.06% |
Reinforcement Learning (RL) | 1 | 3.03% |
Semantic Similarity | 1 | 3.03% |
Semantic Textual Similarity | 1 | 3.03% |
Video Inpainting | 1 | 3.03% |
Automatic Speech Recognition (ASR) | 1 | 3.03% |
Speech Recognition | 1 | 3.03% |