Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
Code for the paper "Language Models are Unsupervised Multitask Learners"
The project huggingface/transfomer has implementation of transfomer based modles (such as GPT-2 ) .
https://openai …
Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
huggingface/transformers has PyTorch implementation of transfomer based models (such as BERT and DistilBERT ).
https://github.com/codertimo/BERT-pytorch
Official Implementation of BERT in TensorFlow
https://mccormickml.com/2019 …
Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
User-based Filtering (Memory-based Filtering)
Item-based Filtering (Content-based Filtering)
Non-negative Matrix Factorization
Neural Matrix Factorization
Variational Autoencoder
Hybrid
New methods like VAE, AE, or Deep Collaborative outperform classical methods like NMF on …
Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
GELU is the best activation function currently (at least in NLP).
,
where \(\Phi(x)\) is the cumulative distribution function of the standard normal distribution.
Things on this page are fragmentary and immature notes/thoughts of the author. Please read with your own judgement!
http://nlp.seas.harvard.edu/2018/04/03/attention.html
https://blog.floydhub.com/the-transformer-in-pytorch/
http://jalammar.github.io/illustrated-transformer/
https://towardsdatascience.com/transformers-141e32e69591