microsoft/onnxruntime

Pull Requests Commits

Document HY-MT2 CUDA benchmark results

Chi Lo committed 18 days ago

c67b9627

Add MatMulNBits dispatch tracing

Chi Lo committed 18 days ago

08c23716

Add Hy-MT2 CUDA benchmark test

chilo-ms committed 25 days ago

b093983c

Fix CUDA plugin config key compilation

chilo-ms committed 30 days ago

a0932465

Clarify Level-1 prepack memory accounting

chilo-ms committed 30 days ago

8a78fb2c

Align CUDA memory estimation with session config

chilo-ms committed 31 days ago

d3cff998

Account for MatMulNBits tactic profiler scratch

chilo-ms committed 31 days ago

5d7d98e2

Clarify Level-1 initialization accounting contract

chilo-ms committed 31 days ago

00d62c9b

Avoid double-counting offline-prepacked GPU weights

chilo-ms committed 31 days ago

44d524ea

Avoid GCC optional initialization warning

chilo-ms committed 32 days ago

2c33db55

Address workspace accounting review feedback

chilo-ms committed 32 days ago

79248e39

Preserve partial profile accounting behavior

chilo-ms committed 32 days ago

a2c9de7e

Account for MatMulNBits prepack memory

chilo-ms committed 35 days ago

661940ae

Report workspace estimate discrepancies

chilo-ms committed 37 days ago

e9e0e3d4

Unify workspace selection state

chilo-ms committed 38 days ago

da899f32

Resolve workspace estimate with resource cost

chilo-ms committed 39 days ago

c8068bce

Clarify initial resource cost calculation

chilo-ms committed 39 days ago

edbd5755

Document pending resource accounting state

chilo-ms committed 39 days ago

6dbbd9b0

Rename size-based resource accountant

chilo-ms committed 39 days ago

a5af386b

Combine profiled and estimated workspace accounting

chilo-ms committed 39 days ago

37560cf0

Report committed workspace estimates after graph partitioning

chilo-ms committed 39 days ago

3afd9306

Wire max-shape inference into workspace estimation (#31613)

chilo-ms committed 39 days ago

Verified 808ba88b

Weightless support for all initializers (#29607)

chilo-ms committed 39 days ago

Verified 53211cd3

[CUDA] Split matmul nbits gemv to multiple cu files (#31837)

tianleiwu committed 39 days ago

Verified 1d69bbd6

[CUDA] Fix CUDA 13 packaging build failures (#31748)

tianleiwu committed 39 days ago

Verified 66eae3e6

[ARM] MLAS: portable machine-code SVE elementwise kernels, FEXPA exp (#31145)

mirounga committed 39 days ago

Verified 1eb1eb96

Fix GatherBlockQuantized default zero point on CUDA (#31693)

Copilot committed 39 days ago

Verified ab7d4609

Make DML runtime fused graph kernel own its model path (#31664)

adrastogi committed 39 days ago

Verified 33d509c7

[Build] Split matmul_nbits cu files (#31834)

tianleiwu committed 39 days ago

Verified 9d52a149

[Build] Reduce nvcc_threads to 1 for linux cuda plugin ci (#31729)

tianleiwu committed 39 days ago

Verified e9c76ade

Older