Where operation can be statically computed when involving splats of
known value. Added handling...

Folds aten::index_select ops under the following conditions:

1. If the input and output are t...

Lowering of torch.aten.all.dim to linalg.

Per PyTorch documentation:

> This function match...

Includes `slice` folder and lit tests

---------

Co-authored-by: Xida Ren <xida.ren.dev@gma...

Just a one line change in a .md file

Link to related RFC:
https://discourse.llvm.org/t/rfc-rename-torch-mlir-compile-apis-and-introd...

A chunk off

https://github.com/llvm/torch-mlir/pull/2856
https://github.com/llvm/torch-mlir/...

Adds an escape hatch from creating a DenseResourceElementsAttr for
single value tensors into De...

If a tensor is initialized by a list with a single constant integer,
this folder turns it into ...

Simple missing scalar operation for multiply floats was missing.

Leaning on the QDQ functionality in torch we can support the QLinearConv
operation by piggyback...

The existing `flatten` lowering did not define what the intermediate
shape was. This could resu...

Signed-off-by: Gaurav Shukla <[email protected]>

So that the CumSum Op in OPT can get the constant that it requires to be lowered to TMTensor

...

`[build]
D:\Dev\iree\third_party\torch-mlir\lib\Conversion\TorchOnnxToTorch\DefaultDomainGtoP.c...

`torch` requires that padding be symmetric for pooling operations. To
support non-symmetric pad...

Fixes bazel build error:
```
ERROR: /root/.cache/bazel/_bazel_root/b89349c08f7224396763d14fe35...

Fix for https://github.com/llvm/torch-mlir/issues/2765

The onnx docs say that you can't do sh...

We do not support average pool when `countIncludePad is set to false.
However if the input is u...

Extracting scalar values from tensors can be implemented via a lowering
to tensor.extract.

With the recent LLVM integrate and changes from
https://github.com/llvm/llvm-project/pull/78260...

To handle the conversion from raw bytes to `DenseElementsAttr` we need
to handle the endianness...

`onnx` explicitly specifies that `raw_data` is stored in `little-endian`
layout. While converti...

Closes nod-ai/SHARK-Turbine#356

Required some massaging of LTC to make it warning clean, and I had to
manually disable some war...

Note that we are waiting for actual FX traced graph support for sparse
tensors. For details see...

This PR adds a check to the CI right after checking out the Torch-MLIR
repository to make sure ...

Linalg has quantized specific operations. We can lower to these
operations when there is a know...

We were seeing some assertion failures after some checks around folders
were tightened up in LL...

Change should be visible :
https://github.com/newling/torch-mlir/blob/docs_update/docs/developm...

See change in test for what the new message looks like.

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

Fixes https://github.com/nod-ai/SHARK-Turbine/issues/352

This allows the following command to be used to ignore sweeping
formatting changes.

```
git...

After noticing a number of commits with unrelated formatting changes,
I think something was cha...

Torch does not have an equivalent matmul operation for integers. Instead
it sidechannels the in...

TypeRange is an ArrayRef<Type> and therefore cannot be safely
instantiated from a list initiali...

Per the RFC and numerous conversations on Discord, this rebuilds the
torch-mlir CI and disconti...

This preserves sparsity at the most obvious places of lowering TORCH
tensors to MLIR RankedTens...

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

As that torch with py3.9 is also used widely.

Support a lowering for dequantization for per channel tensors from
`torch` dialect to a linalg ...

This adds an encoding field to the torch type, using the interfaces for
printing, parsing, and ...

Code used for testing.For the location of CumsumInputDtypeInt32Module in
the repo you can see
...

This includes custom op matching for decomposed operations and fusing
dequantization into dense...

We can plumb the linear matmul into pytorch using its quantized types
with side channel informa...

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

This commit adds mapping from `onnx.pad` op to `torch.pad` op. Currently
it does not support `a...

The logic here is very similar to the conversion for AdaptiveAvgPool1d
#2661 with a few modific...

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

convolution with [time,batch,channel] ordering, as opposed to the
default [batch, channel, time...

https://github.com/nod-ai/SHARK-Turbine/issues/312

Set PyTorch and TorchVision version to nightly release 2024-01-22.

Signed-Off By: Vivek Khand...

Signed-off-by: Gaurav Shukla <[email protected]>

Fixes https://github.com/llvm/torch-mlir/issues/2764

In the case of OPT, there are ConstantO...

https://github.com/nod-ai/SHARK-Turbine/issues/331

Currently transposed convolution is not handled correctly by
`TorchToTosa`. This PR allows tran...

Introduced in 704cfdaf0893fa4078ab744a915a56367282dcaa of @wu-s-john

g++ compiler error:

...

[https://github.com/nod-ai/SHARK-Turbine/issues/328](url)

---------

Co-authored-by: Dave L...

Overly specific docs can get stale easily. It looks like
https://llvm.github.io/torch-mlir/pack...

This commit adds support for gather op in the onnx pipeline.
https://github.com/nod-ai/SHARK-Tu...

Added verification logic to the abstract_interpreter_lib_gen.py

Also made some unit tests

...

Closes nod-ai/SHARK-Turbine#349

We can make the per-tensor version of the operation to the dequantize
operation via marking wit...

We can map the per_tensor case to the `torch.aten.quantize_per_linear`
operation. In this case ...

This PR add torch-to-tosa lowering support for AtenLinalgVectorNormOp

e2e test:
python -m e2...

prims.expand expects the start and end dimensions to be strictly less
than the rank of the tensor.

Implemented ONNX.Range. The spec says the data type for start, limit,
delta are 0-D can be doub...

Co-authored-by: Lisa Liu <[email protected]>

Handles the multiple cases of `onnx` constant values and converts them
to `torch` literal tenso...

- Add fixes for
https://github.com/llvm/llvm-project/commit/af78e5daf0791135485dbd7972ffedb9277...

Handle both `torch.dequantize` and `torch.quantize_per_tensor` including
the op based quantizat...

Add Maxpool ONNX op support.
Add Utils.h/cpp files to create a constant int list for ONNX.

This PR updates the torch-to-tosa conversion with following changes:

- Support torch.none as ...

Closes nod-ai/SHARK-Turbine#290