William Lin f9e3680f11 [perf] Align FastH3 optimized inference profile (#1748)

2026-08-23 02:01:03 -07:00

..

triton_kernels

[bugfix] fastvideo-kernel: fix Triton block-sparse backward logit scaling (bf16 K pre-scaling) (#1730)

2026-08-21 04:50:06 -05:00

__init__.py

[misc] yapf pass under CI's interpreter (3.12) + pin hook language_version (#1702)

2026-08-08 22:24:17 -07:00

block_sparse_attn_256.py

[perf] Add FA4 CuTe backward support for VSA-256 (#1639)

2026-08-19 11:46:49 -07:00

block_sparse_attn_cute_fwd.py

[perf] Add FA4 CuTe backward support for VSA-256 (#1639)

2026-08-19 11:46:49 -07:00

block_sparse_attn_sm100a.py

[perf] Align FastH3 optimized inference profile (#1748)

2026-08-23 02:01:03 -07:00

block_sparse_attn_varlen.py

[misc] yapf pass under CI's interpreter (3.12) + pin hook language_version (#1702)

2026-08-08 22:24:17 -07:00

block_sparse_attn.py

[misc] yapf pass under CI's interpreter (3.12) + pin hook language_version (#1702)

2026-08-08 22:24:17 -07:00

comm_ops.py

[perf] Add fused NVLink all-to-all for Ulysses (#1740)

2026-08-22 23:06:24 -07:00

ops.py

[perf] Add FA4 CuTe backward support for VSA-256 (#1639)

2026-08-19 11:46:49 -07:00

turbodiffusion_ops.py

[misc] yapf pass under CI's interpreter (3.12) + pin hook language_version (#1702)

2026-08-08 22:24:17 -07:00

version.py

[perf] Align FastH3 optimized inference profile (#1748)

2026-08-23 02:01:03 -07:00

vmoba.py

[misc] yapf pass under CI's interpreter (3.12) + pin hook language_version (#1702)

2026-08-08 22:24:17 -07:00

vsa_utils.py

[kernel] sm_100a CUDA block-sparse VSA forward (Blackwell), 64- and 128-token blocks (#1719)

2026-08-20 23:16:25 -05:00