Add arxiv pdf - Vaswani et al 2017 - Transformers
koenrane opened this issue · comments
Proposition to add pdf/link to the following line in the article Why Tool AIs Want to be Agent AIs:
"While LSTM RNNs are the default for sequence tasks, they have occasionally been beaten by feedforward neural networks—using internal attention or “self-attention”, like the Transformer architecture (eg Vaswani et al 2017 or Al-Rfou et al 2018)"
Add this arxiv doc:
Attention Is All You Need
Vaswani is already linked in the previous paragraph so that's unnecessary.