Files
zeekay 26a06b955c perf(oasis): precompute per-forward timestep freqs + causal mask (pitfall #5)
Hoist Tensor::arange (timestep freqs) and Tensor::from_vec (causal mask) out
of the per-forward hot loop to model init, eliminating a CPU-GPU sync per DiT
forward (~310 per 10-step rollout). Parity preserved bit-exact (DiT cosine
1.000000). Also: example now reports per-stage timing (vae-enc/dit/vae-dec).
This is also the precondition for capturing the fixed-shape denoise step as a
CUDA graph.
2026-07-10 16:16:45 -07:00
..