* [MLIR] Add canonicalizer for aten.slice.t op

* Add mlir tests and strength the canonicalize...

Summary of changes:
- Renamed OptionalArrayRefParameter since the name conflicts with an
u...

* Fix as_strided symint

* Re-enable LTC tests

* Re-enable LTC

* Add hardtanh shape infe...

* Add torch-to-mhlo e2e support for aten.gather op

* Add more e2e tests for torch.aten.gathe...

Signed-Off By: Vivek Khandelwal<[email protected]>

* Fix dynamic shapes type verifications

* Add relu6 and binary broadcasts

This adds a very long and obnoxious option to disable crashing tests.
The right fix here is to u...

* Replace empty_like && empty_memory_format with full/full_like

* fix broadcast rank0 tensor

https://github.com/llvm/torch-mlir/issues/1396

Signed-Off By: Vivek Khandelwal <[email protected]>

Co-authored-by: Suraj Sudhir <16977902+s...

Signed-Off By: Vivek Khandelwal<[email protected]>

Co-authored-by: Suraj Sudhir <16977902+sj...

Signed-Off By: Vivek Khandelwal<[email protected]>

Co-authored-by: Vremold <[email protected]>

* Add symint to native functions yaml

* Re-enable LTC

* Fix new_empty_strided and narrow_copy

Strength the shape inference for aten.arange-like op by
1. registering aten.sub and aten.ceil.S...

This PR adds an `AllowedInModuleInitializer` trait to keep track of ops that are permitted in th...

No changes are required to Torch-MLIR, other than updating the MHLO
submodule to point to the b...

Previously we `sudo rm -f` the non-universal zstd installed in the GHA. The CI has this fix but ...

This commit adds bool element type support for `aten.sum` and
`aten.sum.dim_IntList` op.

Signed...

As @oroppas identified, literal strings that are over 16,380 characters
cause the MSVC compiler...

Signed-Off By: Vivek Khandelwal <[email protected]>

Add AtenClampOp conversion pattern to MHLO

Addresses leftover comment from earlier PRs: #1254 , #1265 to remove `torch_dispatch` frontend. ...

TorchLoweringPipelineOptions only applies to the frontend lowering
pipeline.

Summary of changes:
- Updated emitAccessorPrefix since the default value has changed
(http...

It seems as though an upstream change in PyTorch has caused the module
dump to include not just...

This commit adds lowering of `index.Tensor_hacked_twin` op.

Signed-Off By: Vivek Khandelwal <vi...

Signed-Off By: Vivek Khandelwal <[email protected]>

This commit adds lowering of `aten.bitwise_not` op.

Signed-Off By: Vivek Khandelwal <vivek@nod-...

This commit adds lowering of `aten.lift_fresh_copy` op.

Signed-Off By: Vivek Khandelwal <vivek@...

This commit adds lowering of `aten.pow.Tensor_Tensor` op.

Signed-Off By: Vivek Khandelwal <vive...

* Add aten.frobenius_norm.dim op and init its conversion pattern to linalg and MHLO,
* run sym...

The `-DTORCH_MLIR_ENABLE_MHLO` flag was recently introduced for `TorchMLIRTorchConversionPassesI...

Summary of changes:
- Update the dataflow analysis in RefineTypes.cpp
- Add tosa-to-arith pa...

This commit adds decomposition of `aten.linear` op. Due to limited
support at tosa backend in ca...

Reverting commit a6a48ba2338aafdf31bc337f3fb28989b03a155c to revise unit tests and address dynam...

Focuses on statically sized cases such as [2, 3] -> [3, 2].

The ConvertAtenNativeLayerNormOp is delete because we have decomposition already
see https://gi...

* Move CIs to use docker builds

Now that #1234 has landed and anyone can run CI / Release bui...

No idea how this slipped by. Sorry about that.

Fixes #1334

Recording was linking to specific time towards end of recording, changed to remove specific time.

We were defaulting to building Release and running tests. Tests are spawned separately.

* Add decomposition to aten.native_layer_norm

* fix ci error

* [MHLO] add VerifyMhloBackendContract

* guard with macro

We added both ipc=host and explicit ulimits. This _may_ be causing slow downs on GHA. Remove the...

Shows low download count and we can add it back if people ask for it. Should save release artifa...

The new logic has the following benefits:
1. It does not clobber the working tree state. We expe...

Related to https://github.com/llvm/torch-mlir/issues/1227

1. Reduce MHLO #ifdefs
2. Dismiss ...

This is to debug what is causing the exactly ccache look up failures etc.

This should speed up source builds and ccache. May cause issues on macOS (https://github.com/pyt...

- Update MHLO commit to build with LLVM commit hash 00d648bd
- Update TorchToMhlo code to wor...

Caught in the wild here:
https://github.com/llvm/torch-mlir/runs/8046660640?check_suite_focus=tr...

Gets both CI and Release builds integrated in one workflow.
Mount ccache and pip cache as requi...

In the sequential case we weren't sorting, which was confusing.

We use it for more than TorchScript testing now. This is a purely
mechanical change to adjust so...

This ensures that they are always up to date.

This also updates the shape lib to make the new C...

We used to not have "value-semantic" tensors but rather "immutable"
tensors

* Disable LTC by default until upstream revert relands

Tracked with the WIP https://github.co...

Change logic so that we never run the multiprocessing codepath with only
1 worker. That configur...

This commit adds a file explaining the structure of an E2E test, as
well as useful things to ke...

* Add LTC architecture diagram

* Use PNG for diagrams

* Update diagram

* Fix symint related functionalization ops

* Remove zeros xfail from LTC tests

LowerToBackendContract now checks all this consistently.

This commit adds a method to `TestUtils` that generates random integer
tensors with a similar i...

* Build LTC to _mlir_libs directory

* Update CMakeLists.txt

* Revert updating mlir_native_functions.cpp signature, due to https://github.com/pytorch/pytorch...

Now that we depend on MLIR-HLO, we have a slightly more involved process
for updating the LLVM ...

An earlier patch (bb47c166) incorrectly replaced the now-dropped
`OpaqueElementsAttr` with `Spa...

* Resolved Wunused-variable

* Fix Wunneeded-internal-declaration

* Address review comment
...

Also adds a decomposition for aten::conv_transposed2d.input

The use of numbers was arbitrary and was preventing the enum values from
being put in the natura...

* Add decomposition for aten.roll

* add e2e unittest

* refine type of torch.roll

* fix ...

Co-authored-by: Bairen Yi <[email protected]>
Co-authored-by: Jiawei Wu <xremold@gma...

Signed-Off By: Vivek Khandelwal<[email protected]>

We were already hitting many cases where backends different in terms of
the legal ops that they ...

Add MHLO path
Add custom accelarator dialects
Rename Torch Dialect back to original Torch-MLIR...

With llvm/llvm-project@91b6f76, the variable `MLIR_PDLL_TABLEGEN_EXE` is
set as a cache variabl...

Signed-Off By: Vivek Khandelwal<[email protected]>

* Add lowering for _convolution.deprecated