PyTorch implementations of transformer-based language models including BERT, OpenAI GPT, GPT-2, and Transformer-XL with pre-trained models, tokenizers, and utilities for NLP tasks