GitLab

Force einsum to run in fp16

remove leftover usage of DUMMY_INPUTS

Added performance enhancements (XLA, AMP) to examples

Fixed the sample code in the title 'Quick tour'.

Remove unnecessary use of FusedLayerNorm in XLNet

Implement fine-tuning BERT on CoNLL-2003 named entity recognition task

Fix import error in script to convert faisreq roberta checkpoints

the working example code to use BertForQuestionAnswering 

Hey @thomwolf  
This change https://github.com/huggingface/transformers/commit/da26bae61b8c1e741fdc6735d46c61b43f649561#diff-8ddce309e88e8eb5b4d02228fd8881daL28-L29 removed the constant, but one usage of that constant remains in the code.

the working example code to use BertForQuestionAnswering and get an answer from a text and a question

GLUE on TPU

Integration of TF 2.0 models with other Keras modules

Add new BERT models for German (cased and uncased)

Re-order XLNet attention head outputs for better perf