kernel
triton

Commit History

v2: bucketed autotune key next_pow2(T) (v1 re-autotuned per length, 30x slow on variable-length ASR); numerically identical, parity vs eager 5 shapes rel<1e-6
480f425
verified

lee-futo commited on

add LDSA model card (citations, differences, benchmarks)
cd8840e
verified

lee-futo commited on

Uploaded using `kernel-builder`.
47d2a94
verified

lee-futo commited on

initial commit
a98583b
verified

lee-futo commited on