Nicholas Broad

Results 46 comments of Nicholas Broad

deberta v3 uses relative position embeddings which means it isn't limited to the typical 512 token limit. As taken from [section A.5 in their paper](https://arxiv.org/pdf/2006.03654.pdf): > With relative position bias,...

I meant to add to my last post: The max length of 1000000000000000019884624838656 is typically an error when the max length is not specified in the tokenizer config file. There...

Try using `--report_to tensorboard`

What about snapping to word boundaries when highlighting? Currently, users can highlight the middle of words and can miss a character or two if they aren't careful. Maybe a toggle...

Is there anything besides the walrus operator that makes it incompatible with 3.7?

`Protocol` in `typing` is also 3.8

Check what the default and preferred encodings are. Run a python repl and the following lines ```python import sys import locale print(sys.getdefaultencoding()) print(locale.getpreferredencoding()) ``` Compare your MacOS to Docker. I...

https://github.com/google-research/electra#requirements >TensorFlow 1.15 (although we hope to support TensorFlow 2.0 at a future date)

#18674 should fix this. Thanks for reporting!

> I wonder what would be the most breaking. It would be better to have a non breaking approach, but I'm not entirely sure we can get away with it....