Although forward pass works, this has the limitation of not working for
backward pass due to the lack of intermediate tensors needed for
gradient.
Next step is to export a training graph and split it manually
Move CudaKernel from cuda_common.h to a new separate header, cuda_kernel.h. Update include sites to use cuda_kernel.h instead if they need CudaKernel. Inclusions of cuda_common.h are now more lightweight.
Make corresponding changes for ROCM execution provider code.
Other minor cleanup.
* define ordering of reduction across blocks
* save state
* remove debug code
* remove debug code
* review comments
* significant correction for reduction only over blocks on same tensor
* addressing ocmments
* update rocm/lamb.cc to build as well
* remove times 2048*size in multitensor test until threshold error in rocm resolved
* convert tuple => struct as per recomendation
* update comment
* apply perfect forwarding for launch_multitensor to permit passing ref rather than pointer
* remove excess template arguments from rocm lamb.cc launch_multitensor as well
* fixes for AMD build
* pr comments
* run formatter from vscode
* formatter on cuda files
* Introduce VariadicAlias, remove hardcoded alias limits
* Include optional-lite in winml build
Co-authored-by: Sherlock Huang <bahuang@OrtTrainingDev3.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
* ReduceL2Grad and ClipGrad.
* fix win build and amd ci pipeline
* resolve comments.
Co-authored-by: Vincent Wang <weicwang@AiFramework2080ti2.corp.microsoft.com>
* add HSA_NO_SCRATCH_RECLAIM=1 to dockerfile
It is to work around an issue in AMD compiler which generates poor GPU ISA when the type of kernel parameter is a structure and “pass-by-value” is used
* update BUILD.md
* add dockerfile for rocm3.10
* Support to pass initial optimizer states to optimizer graph builder
* Changes for passing init optim state to training session config
* Pass optimizer state through cpp and python frontend
* Cleanup
* Review comments
* Fix windows and mac CI
* Review comments
* review comments
* Review comments
* Frontend review changes
* Fix CI
* save python dictionary to hdf5 representation and load an hdf5 file into a python dictionary
* unit tests for saving data to and loading data from hdf5 file
* Initial running changes
* Checkpointing aggregation changes
* compare with older version
* initial cleanup
* Add zero test, minor fix
* Fix zero test, transform, formatting
* Review comments
* add more unit tests
* review comments
* Try fix CI
* Add additional check on just aggregation code
* Try fix ckpt gen
* Add pregenerated ckpt for CI, enable zero test in e2e
* Moving test to nightly, removing ckpt files
* Add tests to dist GPU CI
* Fix dist test
* Review comments
* Fix test
The implementation of QLinearConv internally does a transpose(NHWC)->im2col+GEMM->transpose(NCHW). This adds a graph transformer to change a model to use a com.microsoft.QLinearConv that supports NHWC natively to avoid unnecessary transposes.
This PR adds infrastructure to automatically cache docker images used in CI builds in a container registry.
Currently, build images are pulled from a container registry for some builds and built every time for others. The container registry requires maintenance to keep the images up to date and building images every time wastes build agent resources.
With this change, a given build image can be looked up in a cache container registry and if present, pulled, and otherwise, built and pushed. The uniqueness of a build image is determined by a hash digest of the dockerfile, docker build context directory, and certain "docker build" options. This digest is part of the image tag in the cache container repository.
The cache container registry will need to be cleaned up periodically. This is not automated yet.
* gradient builder for opset13
* code clean.
* resolve comments
* stop grad for axes input
* add split to stop grad list.
Co-authored-by: Vincent Wang <weicwang@OrtDevTest2v100.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
This commit adds shape inference support for the following ops:
SoftmaxCrossEntropy
SoftmaxCrossEntropyLossGrad
SoftmaxCrossEntropyGrad
LayerNormalizationGrad
Motivation and Context
* Introduce PassThrough op to wait for all gradient ready before weight update
* Compute gradient norm for fp32 runs
* Update FE UT expected value
* Respect enable_grad_norm_clip
* Large model export and run ORT Python support
* Megatron change
refine a bit
workaround self attention issue
use partitioned name for weights when megatron model parallel is enabled
Fix Megatron Transformer Issue (cuased by the renaming)
Add UTs for T5 model parallel
Fix megatron seed issue
fix log a bit
checkkpointing changes + rebase
Unintended reshape transform change
t5 layer norm changes
add t5 layer norm kernel
use template for t5 layer norm
template definition changes
no build error
add CPU cuda kernel
first unit test
other forward unit tests
add T5LayerNormGrad
Add c++ transform and test for T5 LN
minor fix
BART MLP Megatron tranform
Add concat slice transform + test
Cosmetic improvements in concat slice transform
Constant folding bug fix + megatron attention transform for BART
Undo unnecessary changes
* Cleanup
* Remove unnecessary changes
* Cleanup megatron
* Windows build
* Add self attention test graph
* Correcting transforms + cleanup
* review comments
* review comments
* fix build and test failures
* Fix CI
* fix windows CI
Co-authored-by: Peng Wang <pengwa@microsoft.com>
Co-authored-by: Aishwarya <aibhanda@OrtTrainingDev4.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
* cpu send/recv
* clean up send/recv
* remove unused code
* assert and nccl option for mnist
* add build option to enable build with only cpu. Without this, nccl is always enabled which will break build on machine that only contains cpu
* Add USE_MPI distinct from USE_NCCL/USE_HOROVOD
* fix
* fix
* exclude cpu send/recv for machines without mpi
Co-authored-by: Tim Harris <tiharr@microsoft.com>
* Create an Azure Pipeline to merge cpp and python e2e pipelines into one. Still keep cpp 2e2 pipeline until this new pipeline is stable.
Co-authored-by: liqun <liqun@OrtTrainingDev4.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
Some part of code for reduction kernels has been changed in 858040fa,
which cause failures in rocm build since ROCm EP shares some code with
CUDA EP. This PR is to quick fix this failure by not sharing two files
for now to unblock CI enabling on ROCm EP. Another PR for leveraging
858040fa for ROCm EP will be done later.
* Add kernels for AMD GPU.
This PR is mostly about GPU kernels for ROCm EP. Due to similar GPU programming language (CUDA and HIP and similar math library calls, one principle in ROCM EP design is to share CUDA kernels as much as possible for ROCm. Thus, the script amd_hipify.py has been created for converting CUDA kernels to ROCm HIP kernels automatically during compilation phase. But, for some reasons such as perf issue, syntax difference..., some converted kernels need some manual intervention. These kernels will be checked in the repo physically for now. In order to avoid manual intervention, the plan is to refactor CUDA kernels to make them portable between CUDA EP and ROCm EP as much as possible.
Please refer to "HIP Porting Guide" for details.
* like lamb, multi-tensor-apply needs to be disabled for IsAllFiniteOp and ReduceAllL2, current AMD GPU compiler has perf issue for kernel parameter which is a structure with "pass by value".
* Use hipMemsetAsync and add checks on HIP calls.
* move the generated files to build folder.
Co-authored-by: Jesse Benson <jesseb@microsoft.com>
* Split change
* ReduceSum and Split change
* Other op changes, Grad builder, tests, registering required opset 13 ops
* Rebase fixes
* Fix tests, add some more
* Review changes, rebase
* Fix windows build
* Disable new tests for TesnorRT EP
* Disable unsupported for OpenVINO
Co-authored-by: Aishwarya <aibhanda@OrtTrainingDev4.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
