microsoft/onnxruntime

Pull Requests Commits

fix(attention): validate GQA head counts in symbolic shape inference

tianleiwu committed 9 days ago

ced57815

fix(attention): address GQA review feedback

tianleiwu committed 20 days ago

83144876

Keep allowing bias when key and value are 4-D BNSH tensors

tianleiwu committed 21 days ago

d574822b

fix(attention): correct grouped-query caches

tianleiwu committed 23 days ago

d1bd6a98

docs: fix grouped-query attribute placement

tianleiwu committed 23 days ago

555fb892

docs: regenerate contrib operator reference

tianleiwu committed 23 days ago

aff244cc

fix(webgpu): use grouped-query value dimensions

tianleiwu committed 23 days ago

db4e5734

consistent naming

tianleiwu committed 23 days ago

54ba92b7

tianleiwu committed 23 days ago

887b17f7

Replace AttentionParameters::v_hidden_size with hidden size accessors

tianleiwu committed 23 days ago

cbec80dd

Add hidden size accessors to AttentionParameters

tianleiwu committed 23 days ago

c83a7809

Pass kv_num_heads explicitly to ApplyAttention

tianleiwu committed 23 days ago

6da01191

Address review feedback on grouped-query MultiHeadAttention

tianleiwu committed 23 days ago

7688cfe8

address feedbacks

tianleiwu committed 23 days ago

2b43d905

Fix grouped-query shape inference build

Copilot committed 23 days ago

88111b7f

Implement CPU grouped-query attention

Copilot committed 23 days ago

e5c0758a

Finalize grouped-query MHA validation and tests

Copilot committed 23 days ago

51dfb9e6

Complete grouped-query MHA shapes and cache handling

Copilot committed 23 days ago

d09280db

Add grouped-query MHA validation and CUDA path

Copilot committed 23 days ago

fc7c88d8

Validate 'num_scan_inputs' attribute in Scan kernel construction (#31668)

titaiwangms committed 23 days ago

Verified a6e96aa7

[Build] Fix Win Arm64 Cuda Plugin Packaging (#32355)

tianleiwu committed 23 days ago

Verified ac94fed9

[CUDA] Fix ScatterElements reduction dispatch by element type (#29879)

Noperi0r committed 23 days ago

Verified 16e6592b

Validate GatherND indices rank before access (#32349)

apsonawane committed 24 days ago

Verified ef965a9d

Validate ScatterND indexed dimensions (#32345)

apsonawane committed 24 days ago

Verified 01529bd5

optimize int4 prepacking of the weight on CPU (#31690)

xadupre committed 24 days ago

Verified fb6422c4

Fix int32 overflow in softmax_warp_forward offset arithmetic (#32330)

DKAIN-py committed 24 days ago

Verified 57affcea

Gate CPU FP16 Gemm and MatMul on hardware acceleration (#32301)

Copilot committed 24 days ago

Verified d97b5498

[WebGPU] Reject foreign GPU handles in the built-in data transfer (#32317)

4n4ny4 committed 24 days ago

Verified c1f82333

[WebGPU] Vectorize Split when every output segment is vec4-aligned (#32251)

4n4ny4 committed 24 days ago

Verified fe299c0f

Validate SpaceDepth shape arithmetic (#32039)

danielsongmicrosoft committed 24 days ago

Verified 9cca776e

Older