single mention · model architecture
Axial 2D-RoPE
An axial form of 2D rotary position encoding used with non-causal attention in the cited vision encoder.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See positional encoding for the methods filed beside it.
Also called axial 2D-RoPE with non-causal attention.
Filed under model architecture › positional encoding