merge #3680
edgchen1
merged
134 commits into
edgchen1/ort_training_to_master_for_review
from
ort_training_for_merge_to_master
Introduce training changes.
e542cfd0
Enable CI for training.
80dd62a2
Change Tensor::[Set]ByteOffset() to use ptrdiff_t.
3af5a2a2
Add back orttraining-linux-gpu-inference-only-ci-pipeline.yml. (#3182)
fa4dd51e
Initial implementation of graph cut and pipeline
75025461
Merged PR 5686: fix P100/fp16 issues
2f1e997e
Merged PR 5688: Upgrade ONNX submodule to the latest from github ONNX…
2cad08bd
Revert change from RelWithDebInfo to Release in OnnxRuntime.CSharp.sln.
24793f5f
Tweak the dropout calculation.
dc11b829
Update bert-base convergence values
3a7539e0
Udpate License Header (#3212)
4b2c8e88
Fix build issues (#3214)
c5576d70
Register ONNX Training Ops (#3252)
03d14bae
Add ort_training build status file. (#3257)
d82f72e6
Address PR comments (#3255)
61e8a243
Address PR comments (#3256)
ca7985fd
Remove orttraining/tools/scripts/profile directory. (#3268)
d9f628cb
refactor frontend (#3235)
d521efd9
unittests comments (#3278)
5c523322
resolve conflict
ccc3535e
fix build break
9dbc50c4
Make gradient clipping configurable. (#3243)
7f610cac
fix pybind issue introduced by merge
8f7bd51f
Implement pipeline event generator (#3206)
d15c74e7
Aggregated Send/Recv (#3232)
98c28060
Update ort_trainer.py with lazy onnx export (#3244)
6474801c
fix python error
e9877850
Add bias correction in Adam & Lamb for C++ frontend & python frontend…
b38fc0d5
move env to .cc file
9eb792a5
fix windows build
a031af8c
address PR comments (#3312)
66c7579c
Expose frozen_weights in PyTorch Frontend (#3317)
d143b41b
Merge pull request #3285 from microsoft/xuzhu/merge_from_master
0a6ec0df
Addressing PR comments (#3334)
131c65d2
support Huggingface's adamw (#3318)
49e6043d
Implement WhereGrad (#3343)
ffb2a335
Don't cast to fp16 in LayernormGrad (#3328)
1bbc4218
Address PR comments (#3352)
d8f0a0f2
Address master merge PR comments (#3348)
fb2f97a0
Merge remote-tracking branch 'origin/master' into thiagofc/ort_traini…
759818f2
Fix code-base after breaking API changes
83c3da3f
add pipeline graph split script (#3275)
efc8bd73
Fix InferenceSession API
6d769d47
Update Op's Domain and Version (#3356)
614eb438
PR comments (#3374)
2ce90cff
Disable tests (temporary)
28ff88ce
Revert _SliceKernel cuda implementation
0b1e3f1e
Revert Session and InferenceSession implementation
e2afe5e0
Disable GradientCheckerTest tests for GPU/Debug build (#3407)
675035b1
Disable GradientCheckerTest tests for GPU/Debug build (#3407)
d89e5d91
Revert "Addressing PR comments (#3334)" (#3412)
f4376653
Enable loss scale input from Python frontend (#3327)
82c1e1b3
Reapply commit 131c65d; Fix memory regression issue. (#3423)
a3ab2ba0
Disable gradient clipping for E2E test.
95707d22
Merge pull request #3383
15e32b44
View Op - new unit tests and add support for tensor memcpy by offset/…
b3546828
frontend test to use random seed (#3209)
1ddfe124
safeint for region bytes in bfc arena and code clean up (#3447)
f7300848
raid rtol to unblock CI (#3457)
a2985566
Address comments around bfc arena (#3460)
a08f1647
Merge branch 'master' into ort_training
6ba7c99e
Fix onnxruntime_unittests.cmake after merge.
8ea0e596
Fix dynamicslice.cc after merge.
eaa3f652
create pipeline for ci frontend tests (#3422)
e7297e6c
Rename ONNX OPTIONAL to OPTIONAL_VALUE.
84773c61
Get cuda_common.h from master.
0e4080f1
Get onnxruntime/core/providers/cuda/tensor/slice.h from ort_training.
6bbc8095
Get onnxruntime/contrib_ops/cuda/bert/fast_gelu.cc from ort_training.
c5176087
Get onnxruntime/core/providers/cuda/cu from ort_training.
1b465ba8
Get onnxruntime/core/providers/cuda/math/matmul_integer.cc from ort_t…
507d2bb9
Remove FastGelu from activations.
bb2f4279
Fixes for Where, ConcatGrad and ReduceSumGrad (#3415)
03996c7c
Remove orttraining/docker directory. (#3476)
20c7dd9f
Put dropout_default, dropout_random, celu back in the list of broken …
4b5f66ad
fix internal loss scale (#3483)
f5ba9c92
Publish unit test results from Linux and Mac builds (#3480)
cffdff67
Add to list of failing backend tests from master.
571a6d52
Get cudnn_common.cc from master.
7b2fc196
Remove usage of DeviceProp (which is removed in ort_training) from cu…
b670cdc1
Put back SubmoduleCheckoutMode parameter into mac-ci.yml.
bf3df414
Merge pull request #3486 from microsoft/sedymche/merge_master_ort_tra…
5d99f179
Update Graph SetInputs and SetOutputs for training (#3446)
ba7225f9
Fix fp16 type mismatch when graph output is an fp32-only node (#3411)
06b63975
Remove cast to OpKernelContextInternal to get threadpool and directly…
4fa88a0a
MaxBatchSize E2E Test (#3454)
2c7c4507
Add Python API to set random seed: onnxruntime.seed(<seed>)
644bc058
Rename API to onnxruntime.set_seed(<seed>)
2536e806
Address PR comments and clean up. (#3536)
2f16172e
merged with master
80e0c64e
Put safeint_interface include directory into onnxruntime_common inter…
0ec90f70
SoftmaxCrossEntropyLoss-12 forward and backward kernel implementation…
6c1ccb65
Ort training README (#3404)
9fc2b648
Fix GraphTest.UnusedValueInfoSerializes.
be3b9244
Add SafeInt include to WinML targets (#3558)
de543c03
Disable or update flaky tests, improve test random seed accessibility…
2cb8cb81
subgraph type override handling and unit test (#3560)
ca1bbff5
Clean up docs. (#3579)
811bd678
Merge pull request #3557 from microsoft/havenka/master-merge
52cfc98e
Support ONNX test version parsing from path on Windows in onnx_test_r…
f180b71f
Add front-end MNIST test (#3231)
781e1c36
Sync onnx_backend_test_series.py disabled tests (#3603)
ead00f97
Merge remote-tracking branch 'origin/master' into edgchen1/merge_from…
daa14b64
Fix merge issue.
87fad09c
Fix GraphTransformationTests tests.
d50c3e7a
Revert "Convert Gelu to use TryParallelFor (#3599)"
28a0c863
Disable CudaKernelTest.SoftmaxCrossEntropyLoss_LargeSizeTensor becaus…
2e4b9b1d
Add --enable_onnx_tests to Windows builds to allow set up of test dat…
297ab43b
Add --skip_onnx_tests to orttraining Windows builds.
47f1758f
Update Optimizer Domain and Opset (#3602)
d66d5bb8
Remove Windows CUDA 9 build definition and helper scripts. (#3615)
5492d02c
Eliminate Useless Cast during Transformer. (#3606)
d3a2ac5c
Clean up OPTIONAL name conflict workarounds in ort_training. (#3622)
b518cb2a
Merge remote-tracking branch 'origin/ort_training' into edgchen1/merg…
8d09cefa
Merge remote-tracking branch 'origin/master' into edgchen1/merge_from…
8df5076d
Add Lamb shape inference (#3634)
ab70625b
Merge pull request #3607 from microsoft/edgchen1/merge_from_master
bb9b0ba5
Refactoring code related to WARP_SIZE. (#3623)
e4fc8325
fixes for ort_trainer.py to resume from checkpoint (#3510)
0e12d05c
Add check for nullptr in PlannerImpl::FindReusableTensor(). (#3619)
2dd4f7e9
expose training session so the training app could register custom ker…
37f4f743
Expand elimination and Expand gradient. (#3610)
ffe19ae4
Try not to modify base name (#3638)
d9641f29
GatherElementsGrad Kernels (#3627)
6136fd07
Add pipeline transformer for wait/record node (#3513)
f1ba9aaf
fixed mnist bug (#3569)
125f68f3
Simplify and clean code (#3655)
33662480
Change CentOS build to use agent pool because builds on hosted agents…
49a1c5e5
disable broken test in DML (#3666)
ae7da234
Merge remote-tracking branch 'origin/master' into edgchen1/merge_from…
deac4676
Merge pull request #3670 from microsoft/edgchen1/merge_from_master
b4e82913
Merge remote-tracking branch 'origin/ort_training' into edgchen1/merg…
4416d418
Merge pull request #3675 from microsoft/edgchen1/merge_from_ort_training
6ca44e21
Merge remote-tracking branch 'origin/master' into ort_training_for_me…
5a790a4b
Revert "Try not to modify base name (#3638)"
3863bd6f
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub