This results in cleaner IR. In particular, Mlp2LayerModule e2e test has
a dim op that is elimina...

This op is much better behaved than the `torch.tensor.literal` op
(which is the new name of the ...

This removes the dependence of the `torch` dialect on the low-level
builtin types.
Now the `torc...

This replaces the ad-hoc use of `i64` throughout the Torch layer, and
helps to keep it crystal c...

This removes the use of `scf.if`, which required laundering back and
forth between `i1` and `!to...

This fixes a "regression" on ResNet where we weren't folding away all
the control flow. For now,...

This finishes removing the dependence on the basicpy dialect!

Changes:
- Add `!torch.bool` type...

- This removes reliance on basicpy.numeric_constant.
- Also, add OpAsmOpInterface to the `torch....

- Remove dependence on `!basicpy.BytesType`.
- Add `torch.constant.str "s"` analogous to `torch....

This also restructures the docs to a "roadmap" directory, to preserve
previous roadmaps / allow ...

This further eliminates the need for the `basicpy` dependency.

This required adding `torch.prim...

- Make consistent with MLIR Core
- Use `//` or `///` comments.
- Use `bool` type for boolean...

- Add `torch.constant.none` op to construct it (naming is chosen to be
analogous to Torch's re...

Changes:
- MLIR_BINDINGS_PYTHON_ENABLED -> MLIR_ENABLE_BINDINGS_PYTHON
- canonicalizer constant ...

This removes our reliance on the numpy dialect and avoids our off-label
use of the builtin tneso...

This is enough to import the program and get it through the compilation
pipeline. It of course f...

It wasn't printing notes or putting the "error:" in front.

This now gives [much nicer output](https://gist.github.com/silvasean/f048e0f37b04542dae6469b8680...

This code was not exception safe -- it would leave an operation
unattached to anything, which br...

This is a really major and invasive restructuring of the way we get
torch operators (`torch::jit...

Changes:
- representation of arg attributes on functions changed

This allows the canonicalizer to coalesce it like other constants.

This op has complex aliasing semantics, so it is kept mutable for now.

With this, we reduce Res...

Interestingly, TorchScript has its own op (`torch::jit::Operator`)
registry separate from the di...

This trait lets us model the semantics of various aten/torch/numpy ops
that are insensitive to t...

The pipeline is subsumed by our lowering pipelines.

Also add some canonicalizations that finally reduce ResNet down to a
single block.

This is the start of a push to getting ResNet running.

This involves throwing in the towel on a...

It was confusing now that we have `convert-aten-to-std`.

- aten::relu_, aten::max_pool2d, aten::adaptive_avg_pool2d, aten::batch_norm, aten::conv2d

No a...

This removes the need for defining all of the custom propagation logic,
and also adds support fo...

- Rename FrozenRewritePatternList -> FrozenRewritePatternSet

These tests pass on the reference backend.

- Add aten.linear op + shape xfer function + ATen->L...

This inlines global slots if possible. This allows them to participate
in folding, canonicalizat...

Also,
- improve error reporting of e2e framework.

- Move frontend lowering pipelines to c++ (this helps with reproducing
failures in npcomp-opt)...

- add_mlir_doc arg order
- fix some dependent dialects on passes that were now causing errors
- ...

This file needs to adopt best practices for how to structure python
"main"'s, but I don't know h...

This pass verifies that a given module satisfies the contract that we
have for backends. This is...

As described in the code comment:

```
When we import TorchScript IR, we import their entire "co...

The E2E tests can be run with
```
npcpy frontends/pytorch/e2e_testing/torchscript/main.py
```

T...

This is our first op with error semantics, and stresses the system.

There are a few design note...

Recommended review order:
- Changes in frontends/pytorch/examples/
- Changes in python/npcomp/co...

This pass allows shape information to be propagated to return types,
which is nontrivial and can...

This revamps the TORCH_TO_TCF_PASSES to reflect the new layering that we
are doing in the compil...

Currently implemented as a simple intraprocedural dataflow analysis over
a standard ShapedType l...

- ModuleOp no longer has a terminator.

This pass incorporates torch.type_bound info and also removes NoneType
returns (eventually it wi...

The current implementation is just sufficient to do a unary aten.tanh
from the e2e spike, and ju...

This is more consistent and intuitive -- usually the object being
"indexed" or used as a "contex...

These allow users to annotate a known "type bound" on the argument,
which can seed shape/dtype i...

- renames of OwningRewritePatternList -> RewritePatternSet
- also `insert` to `add`
- RewriteP...

* Adds f32 scalar argument support across the ABI boundary.
* Adds support for passing input typ...

We already had the `promoteTrailingOutTensor` flag, but weren't using
it. A inplaceVariantKernel...

This reverts commit e7b96ebefc2ca364428e169daffeaa3804236ec7.

This reverts commit 30a42dea32c8306fa63a652ffe690f8342d5c323.

We should generally be using torch_signature_ods_gen.py for generating
these. Somehow this one s...

Also add `is_write` field.

Signed-off-by: Bairen Yi <[email protected]>

torchvision nightly has not bump to 0.10.0 alpha, so pip installs
torchvision==0.9.0 even with t...

* Import ATen conv2d conversion and test

This is a first attempt at expanding ATen-to-TCF con...

This happens in practice with e.g. ResNet from torchvision (multiple
instances of the same Batch...

Signed-off-by: Bairen Yi <[email protected]>

Signed-off-by: Bairen Yi <[email protected]>

So CI build options are closer to those in `build_tools/install_mlir.sh`.

Also append hash of C...

Signed-off-by: Bairen Yi <[email protected]>

* [torch] Add import support for IValue string Type(s)

* [test] Add test for Strings import

Tracing seems now now capture a 4-operand version of aten::add instead
of 3-operand.

I fixed th...

In terms of IR structure, TorchScript allows types to vary in many
circumstances where MLIR requ...

I could not find a corresponding ListIndex in prim, which seems to
translate to a __get_attr__ u...

This is a funny one. It combines a `for` and `while` loop in one op. We
will need to write some ...

This arises when casting optionals, which happens a lot especially
around handling of default ar...

Used by resnet18.

It seems to originate from a helper `_verify_batch_size`:
[code link](https:/...

This primarily unlocks proper handling of free functions (that is,
functions that are not method...