microsoft/onnxruntime

Pull Requests Commits

[WebGPU] Refactor MatMul algorithm selection (#32630)

qjia7 committed 7 days ago

Verified bde59f36

Bump plugin ep versions (#33214)

tianleiwu committed 7 days ago

Verified 9073c946

[WebGPU] Restore ORT 1.24.4 compatibility with per-Session recording and serialized execution (#33082)

xiaofeihan1 committed 7 days ago

Verified 6e0f154e

Preserve FP16 bias accuracy for unsigned and asymmetric weight quantization

Copilot committed 7 days ago

Verified 7f6e059f

Round FP16 bias-compatible weight scales upward to prevent clipping

Copilot committed 7 days ago

Verified 1bdadfaf

Prevent FP16 QDQ bias-scale underflow with FP32 dequantization

Copilot committed 7 days ago

Verified dcd072b1

Preserve representable float16 scales during integer quantization

Copilot committed 7 days ago

Verified 3e51806b

[CUDA] Single-source remaining GQA preparation scratch sizing (#33118)

chilo-ms committed 7 days ago

Verified 11cba206

[WebGPU] Tile large block-FP8 QMoE expert matmuls (#33075)

sayanshaw24 committed 7 days ago

Verified 29feb3ba

Improve INT4 decode and GEMM profiler buffer safety (#33177)

hanbitmyths committed 7 days ago

Verified e2a99afe

Copilot committed 7 days ago

Verified 1f4ad7ad

Fix model package EP setup and stable API contracts (#32790)

xiaoyu-work committed 7 days ago

Verified 58bacbc1

[WebGPU EP] fixes fp16 accumulation bug in gemm kernel (#33057)

prathikr committed 7 days ago

Verified 0066c380

Add FP16 and BF16 H512 GQA decode support (#33176)

hanbitmyths committed 7 days ago

Verified 426d9b35

Fix empty CPU reductions with negative axes (#33232)

xadupre committed 8 days ago

Verified 9a6089e1

Fix PadFusion changing MaxPool results for negative inputs (#33206)

the0cp committed 8 days ago

Verified e19a9ae8

[WebGPU] Accelerated model loading using Win32 and D3D12 (#33055)

sushraja-msft committed 8 days ago

Verified f289d7a8

[CUDA] Fix QMoE SM120 configuration dispatch on SM121 devices (#33225)

kunal-vaishnavi committed 8 days ago

Verified a477effa

[WebGPU] Reduce QMoE single-token gate sorting to stable TopK (#33058)

sayanshaw24 committed 8 days ago

Verified 8fe86b17

Fix WASM Debug external-data test failures (#33226)

Copilot committed 8 days ago

Verified 2297e770

[CPU] Support mixed types and precision in QuantizeLinear (#32735)

TANGBUDU committed 8 days ago

Verified d7a4f6a3

Make shared WebGPU data transfers stateless (#33208)

mustjab committed 8 days ago

Verified ac44e162

[WebGPU] Guard padded workgroups after dispatch normalization (#33167)

edgchen1 committed 8 days ago

Verified 79d3fcf1

[WebGPU] Fix deferred pipeline callback lifetime after failed waits (#33209)

aciddelgado committed 8 days ago

Verified e379df24

Limit function call depth and fix lint issues (#32339)

apsonawane committed 8 days ago

Verified 17be9c28

Fix WebGPU nightly build (#33162)

edgchen1 committed 8 days ago

Verified 41c59c0f

[Compile API] Add support for writing external initializers and EP context to buffer to enable saving models > 2GB (#31347)

skottmckay committed 8 days ago

Verified 660125b3

Update parameterized requirement from >=0.8.1 to >=0.9.0 (#33022)

dependabot[bot] committed 8 days ago

Verified 06cab16f

Fix isolated provider tests in Windows GPU CI (#33203)

xadupre committed 8 days ago

Verified 864d8712

Bump baseline-browser-mapping from 2.10.31 to 2.11.26 in /js/react_native/e2e (#32509)

dependabot[bot] committed 8 days ago

Verified 17447024

Newer Older