Autoregressive Transformers
GPT
Introduced by Radford et al. in Improving Language Understanding by Generative Pre-TrainingGPT is a Transformer-based architecture and training procedure for natural language processing tasks. Training follows a two-stage procedure. First, a language modeling objective is used on the unlabeled data to learn the initial parameters of a neural network model. Subsequently, these parameters are adapted to a target task using the corresponding supervised objective.
Source: Improving Language Understanding by Generative Pre-Training
Papers
| Paper | Code | Results | Date | Stars |
|---|
Tasks
| Task | Papers | Share |
|---|---|---|
| Language Modelling | 43 | 14.48% |
| Text Generation | 16 | 5.39% |
| Question Answering | 11 | 3.70% |
| Machine Translation | 10 | 3.37% |
| Natural Language Understanding | 9 | 3.03% |
| Text Classification | 8 | 2.69% |
| Sentiment Analysis | 7 | 2.36% |
| Speech Recognition | 6 | 2.02% |
| Self-Supervised Learning | 5 | 1.68% |
Adam
Attention Dropout
BPE
Dense Connections
Discriminative Fine-Tuning
Dropout
GELU
Layer Normalization
Linear Warmup With Cosine Annealing
Multi-Head Attention
Residual Connection
Scaled Dot-Product Attention
Softmax
Weight Decay
