onnxruntime
[CUDA] Add LinearAttentionGate and GatedRMSNorm contrib operators
#31158
Open
Open
[CUDA] Add LinearAttentionGate and GatedRMSNorm contrib operators #31158
tianleiwu
wants to merge
12 commits into
main
from
tlwu/20260730/linear_attention_gates
tianleiwu
force-pushed the
tlwu/20260730/state_window_ops
branch
from
5b6d5dba
to
264c0802
6 days ago
[CUDA/CPU/WebGPU] Add state_window to LinearAttention and CausalConvW…
837e57bd
[CUDA] Add LinearAttentionGate and GatedRMSNorm contrib operators
8456e570
fix(cuda): register bfloat16 LinearAttentionGate/GatedRMSNorm kernels
9df39a68
fix build errors
0671f8e7
Fix test
426d5815
fix test (skip if no cuda ep)
1ce5b08f
Merge remote-tracking branch 'origin/main' into tlwu/20260730/linear_…
c68838cc
Fix CUDA linear attention input validation
741f953c
Avoid gate launch block count overflow
783ab6fb
lintrunner
f3f2ae02
update doc
e0cad08b
Merge remote-tracking branch 'origin/main' into tlwu/20260730/linear_…
5d5d331a
Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub