This model was converted to FP16 from z-lab/Qwen3.8-27B-DFlash2 BF16. >DFlash 2 is a block-diffusion drafter for speculative decoding. It predicts a whole block of tokens in a single pass and keeps the top candidates at every position. A lightweight selector then traces one coherent path through them. Two-tap dynamic convolutions in the backbone keep the draft from decaying toward the end of the block. Decoding is lossless: greedy output matches the target model exactly, and sampling preserves its distribution. «FP16» is M1/M2 Apple Silicon only optimization that leads to a very noticeable prompt processing boost. See «Metal FP32 Vs BF16 Vs FP16 benchmark» and jundot/omlx/pull/880 for details.
Модальности:
Генерация текста
Задача: Генерация текста
Автор: deepsweet
Теги: qwen3, dflash, speculative-decoding, draft-model, text-generation-inference, endpoints_compatible
Лайков: 3 | Загрузок: 195
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
