szl-maskmod

Status

STATUS: import-LIVE on CPU Kernel Hub get_kernel (kernels 0.16.1). GPU attention / Flash / Sage / Flex / Triton stays ROADMAP.

Thing	Label	Method / N / date / what-NOT
Kernel Hub get_kernel	import-LIVE	MEASURED 2026-08-28 2:28pm ET on kernels 0.16.1. HEAD 1e2ebef (1e2ebef3c3faec7bb9ca178e708eb5d7af8ffe69). Legal name szl-maskmod (Python module szl_maskmod). Variants: build/torch-universal (default get_kernel) and build/torch-cpu (backend="cpu"). Working calls: get_kernel("SZLHOLDINGS/szl-maskmod", revision="main", trust_remote_code=True) and the same with backend="cpu". selfcheck ok=true, chain_ok=true, chain_depth=1, max_abs_vs_sdpa_causal=2.384185791015625e-07. What-NOT: no tokens/s; no joules. CPU torch Flex-silhouette (score_mod + block-mask). GPU/Triton not claimed LIVE. Λ = Conjecture 1 (advisory).
GPU attention (Flash / Sage / Flex / Triton)	ROADMAP	Not claimed LIVE. This stamp is CPU torch Flex-silhouette only.

Canonical source: https://github.com/szl-holdings/szl-maskmod

This Hub repo is the publish mirror. ATELIER owns cards. No CUDA benches. Λ = Conjecture 1. Apache-2.0.

from szl_maskmod import maskmod_attn, ReceiptChain, selfcheck
import torch
q = k = v = torch.randn(1, 2, 8, 16)
y = maskmod_attn(q, k, v, causal=True)
print(selfcheck())