microsoft/onnxruntime

Pull Requests Commits

modify orttrainner.py in official file, so will pack into wheel

zhijxu-MS committed 4 years ago

8203261b

no functional changed

zhijxu-MS committed 4 years ago

1798ef54

add modified orttrainer.py

zhijxu-MS committed 4 years ago

a0f3c00f

fix bug in biasdropout

zhijxu-MS committed 4 years ago

7b2482b7

only rank0 will dump tensor

zhijxu-MS committed 4 years ago

b4f0136d

sce gradient doesn't have gpu-fp16 implementation, so also insert cast op for it

zhijxu-MS committed 4 years ago

e19e518a

put sce as fp32 node instead of loss node

zhijxu-MS committed 4 years ago

24854cc2

add code to dump bp graph "bp_graph.onnx"

zhijxu-MS committed 4 years ago

928ce266

convgrad on rc3

Lafi7e committed 4 years ago

5e868e71

fix issue in harding View op's view_count_limit

zhijxu-MS committed 5 years ago

7fbb1c65

put sigmoid/softmax in fp32 mode.

zhijxu-MS committed 5 years ago

a0c7372e

Place shape related compute nodes in CPU (#4940) (#5350)

ashbhandare committed 5 years ago

Verified 38e1bbce

fix quantization of EmbeddingLayerNorm (#5321)

yufenglee committed 5 years ago

Verified 5de47aff

Cherry pick (batch 2) to rel-1.5.1 (#5290)

tianleiwu committed 5 years ago

Verified c00e13a2

Handle missing initializers in allocation planner to fix crashes with DML provider (#5244)

jeffbloo committed 5 years ago

389cca7a

ORT DirectML EP for Iron release, ONNX 1.5 (part 2) (#5263)

fdwr committed 5 years ago

b648fe5f

Fix bug in the back to back quantization of matmul and conv (#5264)

yufenglee committed 5 years ago

eb75b492

bump version to 1.5.1 (#5258)

tianleiwu committed 5 years ago

47447da4

Fix reshape fusion crash (#5252)

wangyems committed 5 years ago

87b15f32

Fix possible ios build break after update to Xcode 12 (#5246)

guoyu-wang committed 5 years ago

fc259de3

Place Shape's output in CPU memory (#5245)

SherlockNoMad committed 5 years ago

9fd76c86

Update BUILD.md training dependency info. (#5240)

edgchen1 committed 5 years ago

9158679c

Update BUILD.md for CUDA versions (#5239)

Changming Sun committed 5 years ago

b9b7c279

update TensorRT docs (#5238)

jywu-msft committed 5 years ago

0cbe240e

Revert to using release SafeInt repo now that it supports a build with exceptions disabled. (#5233)

skottmckay committed 5 years ago

c93f292d

Fix TransposeScaleMatMul and MatMulScaleFusion issues (#5230)

edgchen1 committed 5 years ago

6371ad61

Create profile for all dynamic shape input tensors (#5229)

stevenlix committed 5 years ago

c27f461c

[java] Fixing the buffer semantics. (#5223)

Craigacp committed 5 years ago

4427b1e2

fix _WIN32 (#5218)

jywu-msft committed 5 years ago

c909c677

Update conversion script and process to simplify creating ORT format models and a minimal build (#5217)

skottmckay committed 5 years ago

95b2e316

Older