Add a cuDNN SDPA decode tier to the ONNX standard Attention CUDA kernel (Phase 1) #29715
Initial plan
985ec992
Add cuDNN SDPA decode tier to ONNX Attention CUDA kernel (Phase 1)
6cc5d417
Use named sdpa_kernel constants in cuDNN decode test
57fc923f
copilot-swe-agent
changed the title [WIP] Add a cuDNN SDPA decode tier to ONNX Attention CUDA kernel Add a cuDNN SDPA decode tier to the ONNX standard Attention CUDA kernel (Phase 1) 5 days ago
Address review: drop is_causal gate, assert cuDNN routing, add is_cau…
0490bdc5
Add cuDNN SDPA decode-tier tests (M1-M4) + non-skippable canary
f100dd22
titaiwangms
marked this pull request as ready for review 4 days ago
Add multi-batch nonpad_kv_seqlen rejection test for ONNX Attention (CPU)
34767c6b
Add ort-transformers-gpu-pytest skill (GPU pytest workflow + SDPA dis…
dc1ab7d2
Fix CUDA Plugin EP build: use dual-build-safe GetOrtStream in RunCudn…
df87b1a2
titaiwangms
force pushed
from
275719ed
to
df87b1a2
2 days ago
Login to write a write a comment.
Login via GitHub