onnxruntime
[WebGPU] Add per-head Q/K normalization to PagedAttention
#32808

Merged

Commits

paged attention noncausal webgpu

aciddelgado committed 28 days ago
Update onnxruntime/test/python/transformers/test_paged_attention.py

aciddelgado committed 28 days ago
fix

aciddelgado committed 28 days ago
Merge branch 'aciddelgado/qwen38_webgpu' of https://github.com/microsoft/onnxruntime into aciddelgado/qwen38_webgpu

aciddelgado committed 28 days ago
impl

Prathik Rao committed 28 days ago
copilot

prathikr committed 28 days ago
lint

prathikr committed 28 days ago
docs

Prathik Rao committed 28 days ago
Apply suggestion from @github-actions[bot]

aciddelgado committed 28 days ago
Enforce is_causal attribute values in PagedAttention

aciddelgado committed 27 days ago
docs

Prathik Rao committed 27 days ago
Merge remote-tracking branch 'origin' into prathikrao/gateddeltanet-support-stateupdatecapacity

Prathik Rao committed 27 days ago
Merge prathikrao/gateddeltanet-support-stateupdatecapacity into aciddelgado/qwen38_webgpu

Copilot committed 27 days ago
noncausal local

aciddelgado committed 27 days ago
webgpu paged attention qknorm

aciddelgado committed 26 days ago
Merge main into aciddelgado/paged_qknorm_webgpu

aciddelgado committed 23 days ago
Merge remote-tracking branch 'origin/main' into aciddelgado/paged_qknorm_webgpu

Copilot committed 22 days ago
moe routing record errror fix

aciddelgado committed 22 days ago
add test case

aciddelgado committed 22 days ago
Remove CUDA QMoE repair from WebGPU Q/K-norm PR

aciddelgado committed 22 days ago
Merge main into aciddelgado/paged_qknorm_webgpu

aciddelgado committed 22 days ago
cpu reference test

aciddelgado committed 20 days ago

onnxruntime [WebGPU] Add per-head Q/K normalization to PagedAttention #32808 Merged

onnxruntime
[WebGPU] Add per-head Q/K normalization to PagedAttention
#32808

Merged