TBN / the bittrees news

Briefings

Loading briefing…

Pretraining Transformers with Quantized Softmax in Attention | TBN