This is a much simpler representation than the ad-hoc initializer
function we had before. It is ...

With this, we can import BERT!
```
pt_util ~/tmp/bert.pt --import --exported-name=forward \
| n...

- `module_import -> ivalue_import`, as it mainly tests ivalue_importer.cpp
- `graph_import -> no...

Adds support for lowering a torch.nn.Conv2d module to the Torch Dialect through TorchScript impo...

The first use case is to annotate certain program constructs as either
exported or private. In t...

- dialect registration
- StringAttr::get: order of context arg
- math dialect
- LogicalResult no...

With this + manually setting private visibility on everything, a simple
classifier can be reduce...

This happens in practice. With this, we can globalize slots for the
non-trivial classifier layer...

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

This required restructuring of how we model TorchScript on import. The
main difference is that n...

PyTorch added a Global variable `_is_full_backward_hook` recently.

See https://github.com/pytor...

Modify ACAP_Dispatch to work with latest pytorch
-Remove boxed from convolution's m.impl
-Use re...

This is needed so that output shows up properly in a Jupyter notebook.

For now, we are treating strings as bytes.

This required some careful considerations when defining object identity
for tensors. See the com...

This required some invasive surgery to graph_importer.h/cpp,
specifically moving most of it into...

No changes needed.

It turns out that this was easiest to structure as a general IValue
importer, since torch module...

This allows building NPCOMP as an external project of LLVM, similar to
how CIRCT can be built: ...

Fix more links

Links were broken

These mainly exercise the `module_builder.import_function` function, so
it makes sense for the d...

This silences some warnings I was seeing locally.

- TensorFromElementsOp -> tensor::FromElementsOp
- `cmpi "eq", ...` -> `cmpi eq, ...`. Same for ...

Otherwise `MLIR_BINDINGS_PYTHON_ENABLED=ON` won't work.

- `let typeDesription` -> `let description`
- LLVMIntegerType -> IntegerType

* It was only required with the old python APIs.

* Most updates are mechanical except:
* python/npcomp/__init__.py and python/NpcompModule.cpp:...

Changes:
- linalg init tensor change (outs+init -> just outs)
- IntegerType::get and other built...

Also, touch up the docs.

Point to correct files instead of 404's.

realpath is a GNUUtils package that is not available on recent OSX

TEST=Build on OSX systems ...

--version_script doesn't work on OSX.
Shared libs are .dylibs on OSX.

TEST=Build on iMac Pro...

Fixes:
- use `op->(method on Operation)`
- update for MlirIdentifier in signature of mlirNamedAt...

* This has been anticipated for a long time in that it is quite hard to keep C++ binary compatib...

Fixes:
- renames StandardTypes -> BuiltinTypes
- std.extract_element -> tensor.extract

Date: Mon Nov 30 15:20:30 2020 -0800

Changes:
- finalizing-bufferize is stricter now, and we ...

Date: Mon Nov 30 12:44:35 2020 -0800

Fixes:
- func-bufferize is no longer finalizing, so we n...

This code is really tricky, and was not commented.

Best as I can tell (e.g. from LeakSanitizer), this fixes all the leaks
except for those due to b...

* Going through TODOs on the PyTorch side, this is a big cause of them (not being able to have c...

* Organizes the BasicPyOps.td file by function.
* Renamed `to_boolean` -> `as_predicate_value` (...

* Incorporates changes around LLVM StringRef.
* Ports fix in upstream pybind11 detection.
* Disa...

Although `refCount` is initialized as `std::atomic<int> refCount{0};` in
the definition of Tenso...

* Does not handle all features yet but should conservatively fail on unsupported things.
* Locat...

* Installs numpy as well.

* Incorporates source fixes.
* Uses upstream pybind11 detection logic.
* Patches CI.
* This may ...

Also, update README.md to use the canonical .env file written by
`cmake_configure.sh`.

Note that unlike aten.matmul which has dynamic behavior
depending on the argument ranks (can do ...

- Print out initial PyTorch IR.
- Rename ambiguous "frontend IR" to "TCF IR".
- Add newlines to ...

This reverts commit c60d7b4aae606a99d75b219ae0d29715acb4185d.

It seems to have tickled some sor...

After the recent change of cmake variables
from PYTHON_INCLUDE_DIRS to Python3_INCLUDE_DIRS
an...

This allows invoking TCF to TCP-level conversion more easily, and starts
us towards a path of fa...

It was annoying that we were creating shape.get_extent in the middle of
the bufferization pipeli...

* Days since CMake-is-not-a-language failure: 0

* IREE doesn't have proper install support, so there is some temporary hoaky hacking in our CMak...

This vastly simplifies our code, allowing deleting multiple ops,
simplifying multiple passes, an...

* In most situations, this eliminates the need to explicitly set a path to the Torch cmake files...

Date: Fri Nov 13 15:27:29 2020 -0800

* We aren't quite at e2e parity, but we aren't going back and the old path is bit-rotted.

* Fixes #107
* I wouldn't say I love what had to be done here. Worth a conversation with the PT ...

* A bit gross because I took the chance to upgrade all of the backend bits to the new MLIR Pytho...

* Makes room for the pytorch compiler.
* Some common things can be hoisted from the numpy side b...

This involved adding a `tcp.splatted` op to splat a dynamically sized
init tensor. See rationale...

Date: Wed Oct 28 13:25:48 2020 -0700

- fixup for func syntax change

* We're building libLLVM.so anyway. Saves a lot of time/space to link tools against it.
* MLIR t...

* Incorporates a dep on the new MLIRPublicAPI shared library.
* More work is needed to further s...

* Conversions are very simple, suporting mul, maximum and add (alpha=1 only).
* Example added wi...

* convolution, convolution_backward, _log_softmax, _log_softmax_backward_data, nll_loss_forward,...

Two changes:
- no more "verifyPasses" constructor arg for PassManager
- OpPassManager defaults t...

* Enables -gsplit-dwarf for both LLVM and NPCOMP, reducing the occurrence of the ~GB scale binar...

Now, the only bufferization we have left is lowering tensor constants to
memref, which will hope...

The current code was inserting all build_list ops
after the last constant op since it was assumi...

* This is sufficient to capture the forward and backward pass and gradients of a convolutional m...

- Remove GreedyPatternRewriteDriver.h from files that don't need it
- fix typo shouldBeCloned ->...

- Fixup to OpBuilderDAG
- Update for affine map naming

* None's out Device? args.
* Emits bool tensors if needed.
* Adds some stderr tracing to better ...