Decompose aten._softmax_backward_data into aten math ops. Also decompose
`aten.size` to facilita...

This is to facilitate scalar type conversion in the TorchToLinalg. As
part of adding the helper,...

People might want to lurk on discord, and not even check out the code yet.

This commit adds end to end support for AtenToDtypeOp from aten
to linalg.

Signed-Off-By: Prate...

Add e2e support for torch.pow(Tensor, Float)

Support for returning elemental types. Previously, only
memref types as returning types was supp...

reused clang-format, as changes got deleted

This commit adds decomposition of `aten.Expand` to `aten.BroadcastTo`
op.

Signed-Off-by: Gaurav...

The `aten.log_softmax` is decomposed into `aten.softmax` and
`aten.log` op.

The lowering of `torch.log` op has been added.

Signed-off-by: Prashant Kumar <prashant@nod-labs...

The lowering of `aten.Int.Tensor` op has been added.
The changes has been made as a part of `con...

- This commit adds lowering of `aten.View` to `linalg.TensorExpandShape`.
- This lowering will b...

The types have different levels of categories: where
complex > floating > integral > boolean (> ...

- Changes to dialect conversion that result in no-op materializations
not being created.

- Split out TOSA in the CI.
- Add summary of unexpected test outcomes. This works better when th...

After https://github.com/pytorch/pytorch/pull/65967 the `graph()` method
is only available on `t...

* [tosa] Add Torch.sigmoid fp32 to TOSA

Signed-off-by: Suraj Sudhir <[email protected]>

This commit adds lowering of aten.permute to linalg.generic operation.

Signed-Off-By: Prateek G...

Part of #380

Also
- BoolType is not considered as Scalar
- e2e framework fixes for nan handling...

Lowering of `aten.matmul` op is added from torch to linalg dialect.
The different cases correspo...

* Print more exception info on error during test execution

* Fix formatting

* Add aten::ge...

The pass name changed upstream.

Includes a fix to use `add_mlir_public_c_api_library` for Torch-MLIR's CAPI library, which is no...

Summary:
This commit fixes an off-by-one error in how negative dimensiosn were
being handled in ...

* Also adds a requirements.txt and updates docs to reference it versus stringy pip install.
* Ad...

Keeps redundent pull request and push workflows from running when pushing to branches in the mai...

* Picks up Python configure changes (was pinned to a bad intermediate commit).
* Uses the new m...

- Added a DecomposeComplexOps pass to decompose complex torchOps.
- Refactored `visitAtenArgmaxO...

This brings in https://reviews.llvm.org/D110797. PRs that are in
progress will need to use scrip...

By default CMake is quite happy to use any Python that it finds
to fit the version requirements.

Add argmax lowering from torch to linalg

- Move `run_pipeline_with_repro_report` to a more common place, and use it
consistently
- Atta...

see https://github.com/llvm/torch-mlir/pull/356

We lower through linalg-on-tensors and use RefBackend to run it.
This adds enough support for a ...

Currently I haven't attempted to get the python bits working.
Instead the python bits are forci...

This must be old copypasta.

Need to do manylinux build for it to work across many machines.

This allows `pip` to see them.

This closely follows IREE's
[schedule_snapshot_release.yml](https://github.com/google/iree/blob/...

Will add a CI job that builds and uploads snapshot packages next.

These snuck through.

Replace the original BertSequenceClassification with this new one.
The ops needed to support are...

This is a simple way for externals to plug their backends into the test
suite. They just impleme...

The find/replace mangled it.

This commit (with approval from all contributors) dual licenses
the torch-mlir project under bot...

Use torch tracing to get the module because the original model is not
TorchScriptable out of box.

This brings in the fix for the obscure RefBackend bug we were hitting.

Implement the `lazytensor` python package for converting
lazy computations captured by the Lazy...

TorchFX example has been simplified, since it seems to be hitting that
weird RefBackend bug. Wil...

This leaves no real code outside torch-mlir.

This also renames the "npcomp backend contract" to...

Fix the tests to run with refbackend.

This assertion was reachable from user code (e.g. !torch.qint8 dtype),
which should never be pos...

This moves it into TorchConversion since it is only needed there.

This removes the Backend/ dir...

Accidentally deleted it in the last commit.

Also contains the following changes:
- Remove derefineOp canonicalizer because it's not safe.
- ...

It still had a bit of stale stuff.