flash-attention-with-sink implements an attention variant used in GPT-OSS 20B that integrates a "sink" step into FlashAttention. This repo focuses on the forward path and provides an experimental ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results