Sampler › Denoiser › 12 denoiser blocks › Self-attention › Attention over all positions
Scale and softcap
×no parameters
Divide the scores by √64 and bound them smoothly between −50 and 50.
- in
- [B batch, 12 heads, T queries, T keys]
- out
- [B batch, 12 heads, T queries, T keys]
The scale 1/√64 is the default of flash-attn 2.7.4.post1 and runs inside its kernel, as the cap does (flash_api.cpp 108–121). The library is pinned by its source archive in the review record.
Sources2
Direct evidence
A block normalizes, modulates by the time condition’s shift and scale, projects to queries, keys and values, rotates queries and keys, attends over all positions with softcap, and adds the projected output scaled by a gate; the MLP branch does the same with its own shift, scale and gate.
The implementation, pinned at revision e18c42acdfb7, not yet public
Direct evidence
Diffusion: self-conditioning on, `t_min` 0.05 and `eps` 10⁻⁵; the adaptive time sampler has 100 bins, a minimum weight of 0.0001, an EMA of 0.9 and updates every 5,000 steps; the denoiser’s EMA decay is 0.9999; scheduler `tanh` with d = 3, solver `euler`, `ode_sampling` false, N = 128 by default; the denoiser has 12 blocks of width 768 with 12 heads, time conditioning `adaln`, a standard MLP, softcap 50 and self-conditioning by concatenation; the timestep embedding’s `max_period` is 10.
The implementation, pinned at revision e18c42acdfb7, not yet public