DNeo claims 1.3x faster speculative decoding than DFlash-2
The paper says Mixture-of-Features Attention keeps drafter size and generation quality unchanged.
2 stories, newest first · one card per story
The paper says Mixture-of-Features Attention keeps drafter size and generation quality unchanged.
DCOCO restores intra-block context in parallel speculative drafts, reaching up to 6.6x decoding speedup.
Today 2.6K readers · 2.9K views · 145 reads · all time 206K views