specific method · filed under model architecture
DFlash attention
An attention method named DFlash, listed as a distinct component of the reference inference path.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- core
How sources treat it
One count per evidence span, weakest treatment to strongest.
core 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
The reference inference covers the vision encoder and aligner, DFlash attention, MoE, Hyper-Connections, and the DSpark forward path.
coremodel architecturein DeepSeek-V4-Flash-Vision-ExpDeepSeek
Filed alongside
Other methods under model architecture :: token mixer :: softmax attention.