Roy WangandClaude Fable 5 b74ac0bc61 [Bugfix] Allocate packed outputs in fused_q_kv_rmsnorm so q_b_proj keeps its low-latency GEMM path at decode (#53109) ...

Signed-off-by: yasong.wang <yasong.wang@inferact.ai>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

2026-08-28 02:08:27 +00:00

..

__init__.py

[Bugfix][CI] Prevent common ops imports from initializing CUDA (#50639)

2026-08-01 00:13:29 -04:00

fused_allreduce_rms_norm.py

K3 DSpark AR fusion (#50242)

2026-07-31 11:12:03 +00:00

fused_qk_rmsnorm.py

[Bugfix] Allocate packed outputs in fused_q_kv_rmsnorm so q_b_proj keeps its low-latency GEMM path at decode (#53109)

2026-08-28 02:08:27 +00:00

sequence_parallel.py

[DSV4] Implement Sequence Parallelism (#46789)

2026-08-01 10:22:33 -07:00