This patch adds a new workflow that runs when an update to the
rollpytorch branch by silvasean (in whose name the RollPyTorch action
runs) causes the regular CI build to complete without errors. Upon
execution, this workflow currently just prints the PR number(s) of the
PR created by the RollPyTorch action, but once this is working as
expected, we will add the step to merge the PR changes.
* CI: Skip (redundant) libtorch build when using stable PyTorch version
When we use PyTorch stable builds, there is no need to build libtorch
from source, making the stable-pytorch-with-torch-binary-OFF
configuration redundant with stable-pytorch-with-torch-binary-ON. This
patch drops the redundant configuration from CI.
* CI: Simplify guard conditions for creating and using libtorch cache
Whether libtorch is enabled or not is predicated on a host of conditions
such as the platform, in-tree versus out-of-tree build, and stable
versus nightly PyTorch builds. Instead of repeating these conditions to
guard whether to create or use the libtorch cache artifacts (and getting
them almost incorrect), this patch predicates the relevant pipeline
steps to whether libtorch is enabled, thus making the conditions far
simpler.
The `copy_` op being replaced by `RecomposeSliceCopy_` operates on a
subset of the tensor being mutated, while the `index_put` op being
used to replace the `copy_` op operates on the entire tensor being
mutated. This means that the result type of the `index_put` should be
the type of the input to `index_put` and we need to make sure that
`copy_` does not have users before replacing to avoid type conflicts.
This commit also fixes the result type used for the
`AtenArangeStartStepOp`, and an off-by-1 error when creating the
indices vector.
Lastly, this commit also clamps the `end` value from the slice to the
size of the dimension.
Before inlining a global slot, the users of the global slot are
checked to see if they are `ReadOnly` or `MemoryEffectFree` to make
sure that the global slot is not being mutated. Because the op
`copy.to_vtensor` currently does not have the `ReadOnly` trait, if a
global slot is passed to `copy.to_vtensor`, the pass
`InlineGlobalSlots` will fail.
The op `copy.to_vtensor` is `ReadOnly`, since it does not modify the
contents of the input tensor; it simply makes a new copy. This commit
adds the trait as well as an e2e test that generates the case of a
global slot being passed to a `copy.to_vtensor`.
* feat: split pytorch requirements into stable and nightly
* fix: add true to tests to see full output
* refactor: add comments to explain true statement
* feat: move some tests to experimental mode
* refactor: refactor pipeline into more fine grained difference
* feat: add version differentiation for some tests
* feat: activate more configs
* refactor: change implementation to use less requirement files
* refactor: remove contraints used for testing
* fix: revert some requirement file names
* refactor: remove unnecessary ninja install
* fix: fix version parsing
* refactor: remove dependency on torchvision in main requirements file
* refactor: remove index url
* style: remove unnecesary line switch
* fix: readd index url
Creates a build_linux_arm64 job that builds the release on an arm64 self-hosted runner.
Drop Python 3.10 support
Pass TM_TORCH_VERSION to choose the Stable PyTorch version (since arm64 doesn't have nightly builds)
Borrows nightly / stable Pytorch switch from the WIP
https://github.com/llvm/torch-mlir/pull/2038
When `use_tracing=True` is used to import a model into Torch-MLIR,
several casts get inserted in the IR to bridge the untyped inputs and
outputs with the typed body of the computation. These casts create
extra aliases of tensors that cause the current analysis in
`maximize-value-semantics` to fail.
In particular, the `maximize-value-semantics` analysis assumes that the
only valid alias right after an overwrite is the overwritten
alias. So, if there is a use of a casted version of the overwritten
alias after the overwrite, the analysis fails.
This commit improves the analysis by identifying all cast-like aliases
of the overwritten alias and allowing such aliases to be used after an
overwrite.
Because this issue only arises when using tracing, it cannot be
currently tested e2e, so only lit test is added.
This patch adds a (default-true) input called `cache-enabled` to the
setup-build action, so that when the input is false, ccache is not setup
on the host machine. This patch also sets the input to be false for the
release builds.
* Add AtenIndexTensor StableHlo support
* clean up
* Empty commit, trigger test
* try to debug hanging test
* fix segfulat
* fix bad include
---------
Co-authored-by: zhekun.zhang <zhekun.zhang@bytedance.com>
Since PRs created by the GitHub action bot cannot trigger workflows (and
thus build tests), this patch uses the token for a GitHub app that was
specifically created for the RollPyTorch action.
Lowering torch operations that allow different compatible data types
in its operands to tosa end up generating invalid tosa IR with mixed
data types. In tosa spec, certain operations (generally element-wise
operations) require all operands to have the same data type.
Add wrapper functions for those element-wise tosa ops to perform op
creation with type conversion if necessary.
We previously used a fork of the action/cache repository for the PyTorch
cache since the actions/cache repo did not support read-only caches.
Now that actions/cache supports separate read and write steps, this
patch switches back to the actions/cache repo.
Ashay Rane
Sean Silva
Ramiro Leal-Cavazos
Zhekun Zhang
Yuanqiang Liu
maxbartel
Vivek Khandelwal
Gaurav Shukla
George Petterson
powderluv
Prateek Gupta
github-actions[bot]
TatWai Chong
Maksim Levental
Matthias Gehre