See the documentation in `docs/shape_lib.md` and
`docs/adding_a_shape_function.md` for an overview of the system.
This completely overhauls how we represent shape functions. In
particular, RefineTypes does not infer shapes anymore (only dtypes).
Shape functions are now written in (TorchScript'able) Python.
Recommended review order:
1. Read `docs/shape_lib.md` and `docs/adding_a_shape_function.md`.
1. Code and tests for ReifyShapeCalculations, DropShapeCalculations.
1. Code and tests for SimplifyShapeCalculations.
1. shape_lib_gen.py
1. Code and tests for new RefineTypes pass.
1. Random folders/canonicalizers in TorchOps.cpp and associated test in
`canonicalize.mlir`.
1. New ReadOnly trait inferred from the registry.
1. Any miscellaneous remaining stuff.
Example `-print-ir-after-all` for ElementwiseUnaryModule:
[IR lowering dump](https://gist.github.com/silvasean/e4dc8cbc8d00aac7819602e3cbd8e212).
Example `-print-ir-after-all` for ElementwiseBinaryModule:
[IR lowering dump](https://gist.github.com/silvasean/daf6860ecced732af3568af6b1899113).
This helps keep things organized and also exposes more parallelism to
the build system. It seems though that most of the compile time is
actually spent in the headers though, so the wall time doesn't decrease
as much as I had hoped (and now that the headers are being included
multiple times, the cpu time actually increases a lot, sadly -- will try
to dig into this).
This commit replaces the two rewrite patterns of
maximize-value-semantics with a single pattern that captures the
behavior of both as well as other edge cases previously not
supported. The new pattern works by first performing alias analysis on
a subgraph to see if pattern is applicable, then rewriting all
non-value tensors to value tensors in a single go.
This pass is added to lower ops, which can not be lowered
via the TorchToLinalg pass, such as `torch.bincount` op.
This pass also uses torch-mlir's TMTensor Dialect to lower the
complex ops.
Also add torch.bincount op lowering with the help of TMTensor dialect
Signed-Off By: Vivek Khandelwal <vivek@nod-labs.com>
This commit moves the helper function which are common across
different torch-mlir conversion passes into a common directory
Utils.
Signed-Off By: Vivek Khandelwal <vivek@nod-labs.com>
This commit adds support for integer type inputs for
`AtenMaxOp`, `AtenSumOp`, `AtenSumDimIntListOp`.
Signed-Off By: Vivek Khandelwal <vivek@nod-labs.com>
- This commit adds E2E support for `aten.rand_like` and
`aten.bernoulli_.Tensor` ops.
- The `aten.bernoulli(x)` was implemented as:
`aten.bernoulli(x) = rand_like(x) < 0.5`, assuming 0.5 as default
probability, whereas according to the pytorch documentation:
https://pytorch.org/docs/stable/generated/torch.bernoulli.html#torch.bernoulli
the input x in `aten.bernoulli(x)` is itself a tensor containing
probabilities to be used for drawing the binary random number.
- So this commit fixes the `aten.bernoulli(x)` implementation as:
`aten.bernoulli(x) = rand_like(x) < x`.
- It also fixes the case where the input to `aten.bernoulli_.float` is
an integer tensor. In this case the input must be casted to float type
before passing it as operand to `aten.rand_like` op.
`aten.bernoulli_.float(x, p) = rand_like(float(x)) < p`.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
The view op allows for the new shape argument to have a -1 value for
one of the dimensions, and the op is expected to deduce the size of
that dimension by looking at the sizes of the other dimensions and
comparing it to the total number of elements in the original
tensor. This commit adds this functionality.
This commit does a couple of things. First, it fixes a bug in the
`linalg.generic` body of the `nll_loss_forward` lowering where the
`ignoreIndex` was being compared with the loop index rather than the
current element of the `target` tensor. This was not being caught by
the tests because they were not testing the case where `ingnoreIndex`
actually corresponds to a value in `target`. This has been fixed.
Second, this commit adds support for the `reduction` argument in
`torch.nll_loss_forward` as well as support for 1-D inputs. In order
to simplify the lowering code, I've refactored the code that creates
the `linalg.generic` ops for elementwise and reduction ops into static
functions, to avoid having boilerplate code for indexing maps, etc
that can be very error prone.
Note: The function `convertScalarToDtype` was moved to before all the
conversion patterns, but nothing in it was modified.
This commit adds the invariant to the op `torch.overwrite.tensor.contents` that
both of its operands have the same shape and size. In order to
maintain the invariant, special handling of this op is added to the
`RefineTypes` pass.
This commit adds handling to the `maximize-value-semantics` pass for
the case where a view-like op depends on a tensor that has been
overwritten by a value tensor. The approach for removing the
dependency is to change the input to the view-like op to be a copy of
the value tensor that is being used to overwrite.
This commit also removes `AtenFill_ScalarOp` and
`AtenBernoulli_FloatOp` from the list of view-like ops, since these
ops now have a corresponding op with value semantics into which they
get converted in the `reduce-op-variants` pass.
- This commit decomposes the `aten.batch_norm` op into the
`aten.native_batch_norm` op, instead of lowering it to the
`linalg.generic` op.
- It also adds run-time asserts in the `aten.native_batch_norm` lowering
to make sure that the shape of the weight, bias, running_mean, and
running_var must match the num of features.
- Since the `aten.native_batch_norm` op is not supported at TOSA backend,
all the modules that are dependent on the `aten.native_batch_norm` op
will fail and therefore they should be removed from the TOSA `passing`
set.
- It also moves `checkNotNone` to utility.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
This is intended to explore support for non-structured ops that can't
be modeled by Linalg dialect. `tm_tensor.scan` and `tm_tensor.scatter`
are added as the first such ops. The dialect should aim to be
upstreamed in the future.
This commit adds the op `PseudoAtenFillScalarOp` that represents
`AtenFill_ScalarOp` without the underscore. The approach is the same
as in commit dd998fa4d4.
Adding this op allows for a simpler and more consistent version of the
`empty` and `empty_like` op e2e tests.
- This commit adds lowering of `aten.le.Scalar` and `aten.ge.Scalar` ops
as a part of `convert-torch-to-linalg` pass.
- It also creates a new test script `elementwise_comparison.py` for all
element-wise comparison ops.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
This commit adds the op `PseudoAtenBernoulliFloatOp` that represents
`AtenBernoulli_FloatOp` without the underscore. This is needed to make
sure that the `ReduceOpVariants` pass turns the in-place op into an op
that takes value tensors as inputs, otherwise the
`MaximizeValueSemantics` pass will not be able to add value semantics
correctly.
- This commit adds lowering of `aten.eq.int` op as a part of
`convert-torch-to-std` pass.
- It also refactors the code for binary comparison ops lowering.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
- This commit adds lowering of `aten.Bool.Tensor` and
`aten.Float.Tensor` op as a part of `convert-torch-to-linalg` pass.
- It also adds support for returning bool types.
- It also fixes lowering of the `aten.Int.Tensor` op for non-zero rank
input tensors.
- If a scalar number is converted to a 0-d tensor and passed on to the
`aten.Float.Tensor` op, it folds to the scalar number.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
Some of the lowerings use the result type obtained from the op itself
to tell the `linalg::GenericOp` what the type of the result should be
rather than using the type of the result tensor given to the
`linalg::GenericOp`. This becomes a problem when the result type of
the op has static size information and the result tensor used in
`linalg::GenericOp` has dynamic dimensions, for `linalg::GenericOp`
expects the result type to be equal to the type of the output tensor.
This commit replaces the use of the result type from the op itself
with the type of the result tensor passed to `linalg::GenericOp`.
In order to not create too many dynamic/static versions of the same
e2e test, e2e tests have only been added to the ops that currently
fail when used with static sizes.
* [tosa] Support for AtenNe[Tensor|Scalar]Op, AtenLog2Op,
AtenBitwiseAndTensorOp, AtenSquareOp and AtenThresholdOp
* Fix for Issue #532 - Mixed input types for few ops and updated few
tests to use i32 instead of i64
Signed-off-by: Anup Gangwar <anup.gangwar@arm.com>
Co-authored-by: Anup Gangwar <anup.gangwar@arm.com>
This commit fixes an error in the refine types pass of constant
allocation ops. The function used to set the dtype,
`fillInDtypeGivenDtypeAndDataType`, takes two torch types as arguments,
but a torch type and a standard MLIR type were being passed into it.
This commit also fixes the way the dtype was calculated in
`visitAtenToDtypeOp`. This op was also passing a standard MLIR type as
an argument to the `fillInDtypeGivenDtypeAndDataType`
function. Moreover, since the op `aten.to.dtype` has the dtype
argument as not optional, all that is needed is to match
against the int value to extract the dtype.
- This commit adds `aten.assert` op in the Torch dialect.
- The `aten.assert` op is lowered to `mlir::Assert` op.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
- This commit adds support for `aten.native_batch_norm` operation.
- The current implementation only supports inference mode of
`aten.native_batch_norm` op.
Signed-Off-By: Gaurav Shukla <gaurav@nod-labs.com>
The lowering of aten::nll_loss_backward op has been added
from torch to linalg dialect. The changes has been made as
a part of -torch-convert-to-linalg pass.
Signed-off-by: Prashant Kumar prashant@nod-labs.com
This PR include the following pieces:
- Add torch `Generator` type. `Generator` type is converted to i64 in
refbackend type converter.
- Add seed managment support for the default global generator.
`torch_c.getNextSeed` op is used to get the seed. On refbackend, the
`torch_c.getNextSeed` is lowered to load/store from [0] of global
variable `default_generator` memref<i64> in `InsertRngGlobals` pass.
- Add `aten.uniform_` and testing as an example op for RNG ops. Add
`torch.pseudo.aten.uniform` op. It has the same operands and return as
the `aten.uniform_` from the op registry except for value semantics.
The added e2e maxpool testcase from #545 was not getting a static shape
due to an unfolded prim.If when RefineTypes was called. This was because
of unfolded torch.iaten.__is__ and torch.prim.unchecked_cast operators
with torch.derefine operands.
* [tosa] Support for AtenCeilOp and AtenReciprocalOp
* [tosa] Support for comparator ops, Aten[Gt|Lt|Eq][Tensor|Scalar]Op with scalar constant
* [tosa] Support for Scalar variants of Aten[Mul|Div|Add|Sub] Ops with scalar constants
Signed-off-by: Anup Gangwar <anup.gangwar@arm.com>
Co-authored-by: Anup Gangwar <anup.gangwar@arm.com>
- Common code as TF repository, being moved to MLIR core.
- Will support further legalizations to be published.
Signed-off-by: Suraj Sudhir <suraj.sudhir@arm.com>
Note that to enable folding of the code coming from an example
like the ConstantPad2dStaticModule e2e test, support for other
operations had to be added/improved:
- aten::neg.int
- aten::eq.float
- aten::eq.str
- prim::Uninitialized
This commit adds lowering of `aten.threshold` op
This commit adds lowering of `aten.threshold_backward` op
Signed-Off By: Vivek Khandelwal <vivek@nod-labs.com>
This involes the following 2 parts:
- Change refine type to propagate more static shape info.
- Get as much static shape info as possible when creating the result
tensor when converting to linalg.
- This commit adds E2E support for `aten.ones_like` and
`aten.zeros_like` ops.
- Adds support for non-None `dtype` argument of `aten.empty_like` op.
- All the unit test cases related to constant tensor allocation like ops
are moved to a different file named `constant_alloc.py`.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
This commit adds lowering of `aten.arange.start_step` op.
This commit decomposes `aten.arange` and `aten.arange.start` into
`aten.arange.start_step` op.
Signed-Off By: Vivek Khandelwal <vivek@nod-labs.com>
- It folds `aten.to.dtype` when the input tensor type and result type
are exactly same.
- It folds `aten.view` when the rank of both the input tensor type and
result type is unity.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
We only handle the expanding OR collapsing cases, we do not handle
expanding And collapsing happening at the same time or cases where
it's neither collapsing nor expanding like view of [2,3] for
3x2 tensor.
It's assumed that if a shape list element is got from
`aten.size(tensor, dim)` the corresponding dim is not splitted or
collapsed. This assumption makes it easier to deal with dynamic shapes.
- Added E2E support for `aten.eq.Tensor` and `aten.lt.Tensor` ops. Both
the operands are expected to be of the same type, i.e., type promotion
is not addressed as a part of this commit.
- Added E2E support for `aten.eq.Scalar` and `aten.lt.Scalar` ops.
Tensor operand type to Scalar operand type promotion has not been
handled in this commit.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
The existing implementation of `ConvertConstantTensorAllocOp<>` requires
a C++17 feature `if constexpr ()`. This commit removes the use of that
feature to support the implementation even for lower C++ versions.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
Add the required lowerings and correct test cases.
These op produce zero-d tensors and it was incorrectly mentioned in
refine types to produce 1d tensor of size 1.
- Templatize `aten.zeros` and `aten.ones` ops lowering.
- Add E2E support for `aten.empty` op.
- Add Integer type support in `aten.mul.Scalar` op lowering.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
`aten.gt.Tensor` op has been added in torch dialect and the
lowering of the op has been done to the linalg dialect.
Signed-off-by: Prashant Kumar <prashant@nod-labs.com>
This commit adds support for aten.native_layer_norm operation. Here
the previous code for aten.layer_norm is tweaked a little bit to
accomodate both mean and variance values alongwith the layer norm
value. This commit also adds decomposition of aten.layer_norm into
aten.native_layer_norm, which was previously getting lowered directly
to linalg.
Signed-Off-By: Prateek Gupta<prateek@nod-labs.com>
This commit adds lowering of `aten.squeeze.dim` op into
`linalg.TensorCollapseShape` op. Here, the dim(th) dimension of the
input tensor is not supposed to be dynamic.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
This commit adds lowering of `aten.gt.Scalar` and `aten.where.self` as a
part of element-wise ops lowering.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
Support for passing memref of bool types as a function argument
and return is added in ref-backend.
Signed-off-by: Prashant Kumar <prashant@nod-labs.com>
The op lowering has been added as a part of `torch-lower-to-linalg`
pass. This takes care of ignore_index but the weight and reduction
operand is still to be accounted for.
Signed-off-by: Prashant Kumar <prashant@nod-labs.com>
- Supports variants with multiple dims, one dim, all dime
- Leverages legalize_common and legalize_utils code from
TensorFlow-TOSA work
Signed-off-by: Suraj Sudhir <suraj.sudhir@arm.com>
There is an op name change that requires trivial changes.
Also, some of the warning has been fixed.
Signed-off-by: Prashant Kumar <prashant@nod-labs.com>
Many reduction ops take as an argument an optional output dtype that
can change the type of the input tensor before the reduction is
performed. This commit adds support for the optional dtype flag that
had been previously ignored.
Test:
/tools/torchscript_e2e_test.sh -f 'ReduceSumDtype'
/tools/torchscript_e2e_test.sh -f 'ReduceSumDImIntListDtype'
This commit adds lowering of `aten.Squeeze` op into
`linalg.TensorCollapseShape` op. The size 1 dynamic dimensions are not
handled as a part of this commit.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
This is to fold the common pattern from Bert inference like:
```
%111 = torch.prim.NumToTensor.Scalar %110 : !torch.int ->
!torch.vtensor<[],si64>
%112 = torch.aten.Int.Tensor %111 : !torch.vtensor<[],si64> ->
!torch.int
```
The lowering of aten.fill.Scalar has been added.
The changes have been made as a part of -torch-convert-to-linalg pass.
Signed-off-by: Prashant Kumar <prashant@nod-labs.com>
This commit fixes a type promotion bug when NumToTensor was given a
float as an argument. In particular, the rules for type promotion of a
scalar vary depending on if the scalar is part of a tensor op or
not. NumToTensor falls under the second category, but it was being
treated as part of the first category.
aten.log_softmax_back_data op lowering and required
tests has been added. Some NFC have also been added.
Signed-off-by: Prashant Kumar prashant@nod-labs.com
This commit adds lowering of `aten.mul.Scalar` and also adds
decomposition of `aten.addmm` to `aten.mul.Scalar`, `aten.add.Tensor`
and `aten.mm` ops.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
Now, aten::linear supports rank 3 inputs. This is a fix
for upcoming bert-inference task. The correct way should be
to support broadcasting in `aten.matmul` op and decompose
`aten.linear` into right ops.
This commit adds new operation `aten.gelu_backward` in the aten
dialect and adds lowering of this operation from aten to linalg.
Signed-Off-By: Prateek Gupta <prateek@nod-labs.com>
This change is to unblock the work of some backprop ops returning more
than one tensors. We will need to think of a more scalable approach
in the future if more flexible return types combinations are needed.
- Remove use of conversion construction macros
- Add mul and div op conversions
- Add corresponding tests
Signed-off-by: Suraj Sudhir <suraj.sudhir@arm.com>
This is to facilitate scalar type conversion in the TorchToLinalg. As
part of adding the helper, this PR also:
- Updated `AtenAddTensorOp`, `AtenSubTensorOp` to use the helpers to
support more type variants.
- Added e2e type promotion testing.
- Added i32 memref return/arg type to support e2e testing.
Support for returning elemental types. Previously, only
memref types as returning types was supported. All the hacky ways
to write tests which return elemental types should be taken care of.
Signed-off-by: Prashant Kumar <prashant@nod-labs.com>
The lowering of `aten.Int.Tensor` op has been added.
The changes has been made as a part of `convert-torch-to-linalg` pass.
Signed-off-by: Prashant Kumar <prashant@nod-labs.com>
- This commit adds lowering of `aten.View` to `linalg.TensorExpandShape`.
- This lowering will be successful only when one or more static
dimensions are expanded.
- It also fixes a typo in `ConvertAtenFlattenUsingIntsOp` conversion
pattern.
Signed-Off-by: Gaurav Shukla <gaurav@nod-labs.com>
The types have different levels of categories: where
complex > floating > integral > boolean (> means left hand
side has higher category).
The operands have different levels of priorities where:
dimensioned tensor > 0-dim tensor > scalar == wrapped 0-dim tensor.
This is represented by the `ResultTypeState.dimResult`,
`ResultTypeState.zeroResult` and `ResultTypeState..wrappedResult` in
the source code.
For operands of the same priorities, the result type should be the
highest categories with sufficient width to hold all operands.
By default, only the highest priority operands participate in the type
promotion logic. Lower priority operands participate if they are in
a higher category than any higher priority operands.
For example, <[],f32> (lower priority) and <[1], si64> tensor would
result in <[?],f32> tensor because floating > integeral. Another example
<[],f64> (lower priority) and <[1], f32> tensor would result in
<[?], f32> tensor because f32 and f64 are the same category.
The ScalarType enum definition, type promotion table, ResultTypeState
struct definition and some helpers are copied from
aten/src/ATen/native/TypeProperties.*
Other references:
- https://pytorch.org/docs/stable/tensor_attributes.html#type-promotion-doc
- https://github.com/pytorch/pytorch/issues/9515
Other minor changes:
1. Fix `visitExpandLikeOp` to consider cases where the given sizes list
size is larger than the input rank.
2. Add back the somehow deleted `torch.aten.softmax.int` tests in
decompose-complex-ops.mlir.
Part of #380
Also
- BoolType is not considered as Scalar
- e2e framework fixes for nan handling
- `tu.rand(..., low=, high=)` support
- delete unused variable (fix warning)
- Add IouOfModule from #380 to e2e test suite (this is a common
calculation in vision models)
Your branch is ahead of 'origin/main' by 1 commit.
Lowering of `aten.matmul` op is added from torch to linalg dialect.
The different cases correspond to
https://pytorch.org/docs/stable/generated/torch.matmul.html.
TODO: Broadcasting in case of batch-matmul is yet to be taken care of.
Signed-off-by: Prashant Kumar <prashant@nod-labs.com>
* Print more exception info on error during test execution
* Fix formatting
* Add aten::gelu lowering
Co-authored-by: Boian Petkantchin <boian@nod-labs.com>
Includes a fix to use `add_mlir_public_c_api_library` for Torch-MLIR's CAPI library, which is now required (note: upstream sample has it the right way).
Disabled a TOSA test per discussion: https://github.com/llvm/torch-mlir/issues/379
Summary:
This commit fixes an off-by-one error in how negative dimensiosn were
being handled in the lowering of transpose. This commit also adds
tests to transpose and unsqueeze to test negative dimensions.
- Added a DecomposeComplexOps pass to decompose complex torchOps.
- Refactored `visitAtenArgmaxOp` and `visitAtenAnyDimOp` to
`visitReductionAlongDimIntOp`.
- Moved some helper functions into
torch-mlir/Dialect/Torch/Utils/Utils.h to be shared by multiple files.
- Added support for f64 tensor as argument and return types.
We lower through linalg-on-tensors and use RefBackend to run it.
This adds enough support for a "tanh" op. Adding more ops should be
fairly mechanical now that things are wired up. Run with:
```
./tools/torchscript_e2e_test.sh -c tosa
```
The backend structure is very similar to linalg-on-tensors based E2E
backends and is a nice parallel (see `tosa_backend.py`). Actually, this
forced a nice refactoring to the layering here. We removed
`torchscript-module-to-linalg-on-tensors-backend-pipeline` and instead
require separately running
```
torchscript-function-to-torch-backend-pipeline,torch-backend-to-linalg-on-tensors-backend-pipeline
```
This highlights the step that lowers to the "torch backend contract"
of cleaned up `torch` dialect ops is a critical step in the lowering.
Going forward, that is the key load-bearing contract of the torch-mlir
project, not the linalg-on-tensors backend contract.
Recommended review order:
- `TorchToTosa.cpp` / `TorchToTosa/basic.mlir`
- `python/torch_mlir_e2e_test/torchscript/configs/tosa_backend.py` and
the new `utils.py` file there.
- `python/torch_mlir_e2e_test/tosa_backends/linalg_on_tensors.py` and
`abc.py` in that directory for the TOSA backend e2e interface.
- other misc mechanical changes
This commit (with approval from all contributors) dual licenses
the torch-mlir project under both the standard LLVM license and the
standard PyTorch license. This will facilitate moving code between
torch-mlir and the two upstream projects.
The standard file comment is now:
```
// This file is licensed under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
// Also available under a BSD-style license. See LICENSE.
```
See `LICENSE` in the project root for the terms of both licenses.
Implement the `lazytensor` python package for converting
lazy computations captured by the Lazy Tensor Core into MLIR.
This PR also fixes a few things with `torchfx` and its example
Also contains the following changes:
- Remove derefineOp canonicalizer because it's not safe.
- Support for optional tensor and list tensors in reduceOpVariant. This
only works for some special detected and easy to handle cases. For list,
it covers the case list is got from a `ListConstruct`. For optional, it
covers the case optional is constructed from a `DerefineOp`.
- Remove the `inferReturnTypes` for `FromBuiltinTensorOp` because it's
not safe to deduce types from the input. For example, a built-in tensor
of i8 could be converted to si8 or ui8. It's better to let the user
specify the return type explicitly.
A few remain in examples/docs that will be naturally be updated in due
time.
This regresses the list support and the general direction of more widely
supported control flow, lists/dicts/globals that we were going for with
the TorchScript path. The idea is that we are deferring that work to
make torch-mlir a very clean standalone thing. We will reboot it,
probably using some of the tools of iree_pydm to make it simpler, and in
a more natural place (such as an iree-torch repo that depends on IREE and
torch-mlir to build a working PyTorch frontend solution for IREE -- it
was really weird that npcomp depended on IREE).
`tools/torchscript_e2e_test.sh` is all green.
This needs a few passes I put into torch-mlir/lib/RefBackend (not to be
confused with `npcomp/lib/RefBackend`, which will soon be deleted).
For the sake of review, since this brings together a lot of things, I
split this into its own commit. I temporarily commented out some "list"
stuff that we are going to remove as part of the torch-mlir refocus.
Our new dependency management solution relies:
- on the C++ side with the public iree-dialects project, which we
include and are using as representative of some missing upstream
ops (so we treat them "as if" they were upstream, with the hope of
upstreaming them after some codevelopment has happened)
- on the Python side, with simple PYTHONPATH manipulation or installed
Python packages. No CMake stuff required.
This moves the bulk of the Python code (including the Torch interop)
from `frontends/pytorch` into `torch-mlir/TorchPlugin`. This also
required reconciling a bunch of other Python-related stuff, like the
`torch` dialects.
As I did this, it was simpler to just remove all the old numpy/basicpy
stuff because we were going to delete it anyway and it was faster than
debugging an intermediate state that would only last O(days) anyway.
torch-mlir has two top-level python packages (built into the
`python_packages` directory):
- `torch_mlir_dialects`: `torch` dialect Python bindings (does not
depend on PyTorch). This also involves building the aggregate CAPI for
`torch-mlir`.
- `torch_mlir`: bindings to the part of the code that links against
PyTorch (or C++ code that transitively does).
Additionally, there remain two more Python packages in npcomp (but
outside `torch-mlir`):
- `npcomp_torch`: Contains the e2e test framework and testing configs
that plug into RefBackend and IREE.
- `npcomp_core`: Contains the low-level interfaces to RefBackend and
IREE that `npcomp_torch` uses, along with its own
`MLIR_PYTHON_PACKAGE_PREFIX=npcomp.` aggregation of the core MLIR
python bindings. (all other functionality has been stripped out)
After all the basicpy/numpy deletions, the `npcomp` C++ code is now very
tiny. It basically just contains RefBackend and the `TorchConversion`
dialect/passes (e.g. `TorchToLinalg.cpp`).
Correspondingly, there are now 4 main testing targets paralleling the
Python layering (which is reflective of the deeper underlying dependency
structure)
- `check-torch-mlir`: checks the `torch-mlir` pure MLIR C++ code.
- `check-torch-mlir-plugin`: checks the code in `TorchPlugin` (e.g.
TorchScript import)
- `check-frontends-pytorch`: Checks the little code we have in
`frontends/pytorch` -- mainly things related to the e2e framework
itself.
- `check-npcomp`: Checks the pure MLIR C++ code inside npcomp.
There is a target `check-npcomp-all` that runs all of them.
The `torch-mlir/build_standalone.sh` script does a standalone build of
`torch-mlir`.
The e2e tests (`tools/torchscript_e2e_test.sh`) are working too.
The update_torch_ods script now lives in
`torch-mlir/build_tools/update_torch_ods.sh` and expects a standalone
build.
This change also required a fix upstream related to cross-shlib Python
dependencies, so we also update llvm-project to
8dca953dd39c0cd8c80decbeb38753f58a4de580 to get
https://reviews.llvm.org/D109776 (no other fixes were needed for the
integrate, thankfully).
This completes most of the large source code changes. Next will be
bringing the CI/packaging/examples back to life.
This creates the `external/torch-mlir` directory as an
LLVM_EXTERNAL_PROJECTS-compatible project (analogous to
`iree-dialects`) and completes movement/rename of all pure MLIR C/C++
compiler code into there. The next step will be to move all the Python
code / code that links/includes PyTorch C++ code (which currently lives
in `frontends/pytorch`) into a subdirectory here.
I call this "earthmoving" because it is mostly mechanical changes and
renames. As a quick summary (we can change this down the road easily)
- C++ `mlir::NPCOMP::Torch -> mlir::torch::Torch`
- CAPI `npcompTorchListTypeGet -> torchMlirTorchListTypeGet`
- preprocessor `#ifndef NPCOMP_ -> #ifndef TORCHMLIR_`
- CMake `NPCOMPFoo -> TorchMLIRFoo`
The goal of this is to create a standalone project creating a center of
mass for entry into the MLIR ecosystem from PyTorch, suitable in scope
for eventual inclusion/ownership in PyTorch. The idea is that
`external/torch-mlir` will some day be pulled out into its own
repository, and then npcomp will simply pull it in as a submodule.
Layering-wise, what lives in `torch-mlir` lowers code from PyTorch
(currently TorchScript, but TorchFX or pytorch/xla-style tracing are
possible extensions) down to what we have been calling the "Torch
backend contract" which is cleaned up IR (inlining, simplifcation,
conversion to value tensors, ...) entirely in the `torch` dialect. This
is the branching off point for further lowering, of which npcomp takes
one opinion (outside `torch-mlir` of course!), namely the
`TorchConversion` dialect/transforms which lower to IR suitable for IREE
and other linalg-on-tensors based lower-level compilers.
Summary of changes:
- move `{include,lib,test}/Dialect/Torch` into `torch-mlir`
- move relevant parts of CAPI into `torch-mlir`.
- leave a few things related to the `torch-mlir` Python build commented
out, which should be resolved in a subsequent change.
This plumbs through a vertical slice of support for lists.
The main chunk of new code here is AnnotateABIPass which captures the
program signature at the Torch backend contract layer, right before we
start `TorchConversion`. The `TorchConversion` lowering process is lossy
w.r.t. types, so it's necessary to do this for all targets in general.
Like using `!iree.list` directly, we use IREE's ABI annotation
representation for this, although there is nothing very IREE-specific
about it (see
https://github.com/google/iree/blob/main/docs/developers/design_docs/function_abi.md)
We change `ListLiteralModule_basic` to use `!torch.int` because IREE
doesn't support f64 yet (and we don't yet have a way for users to say
that they want `!torch.float` to lower as f32).
Recommended review order:
- AnnotateABIPass and tests
- Arg marshaling in npcomp_backend.py and `iree.py`
- Updates to `list_programs.py` / `xfail_sets.py`
- Moving DeleteDeadIREEListsPass to Backend/Common, so that backends
that don't support lists can use it. RefBackend uses that pass, for
example.
This contains the following changes:
- Fix optional knowledge propagation. The initial knowledge should
always be NotNone for the operations we implemented.
- Add Folder for `prim.dtype`
We were not filling the `outs` with the neutral element of the
reduction, which resulted in reading uninitialized values (we were
getting lucky that sometimes the uninitialized buffers were all zero's).
Also,
- Slight tweak to error messages in the e2e framework.
- builder.getSymbolRefAttr is gone.
- OpAsmOpInterface's getAsmResultNames method needs explicit override
- a bunch of churn for builtin.func needing to be made explicit (and
sometimes implicit?)
- operation printers no longer need to print the operation name
themselves.
- snuck in beneficial trivial addition to TmpDeleteDeadIREEListsPass to
test a particular upstream change e2e with my local patchset.
It doesn't work properly with the new dialect registration framework.
This was latent and only was exposed when running through npcomp-opt.
Not worth investing the brainpower to fix now.
- Add `!torch.optional` knowledge tracking
- Changes to improve type propagation for branches and terminators. See
examples in `refine-types-branch.mlir`
- Refator to separate handling of different ops from `visitOperation`
- Add refine types for a few new ops
This converts a basic list op (torch.prim.ListConstruct) to the IREE
dialect.
```
def forward(self, x: float):
return [x, x]
```
turns into:
```
builtin.func @forward(%arg0: !torch.float) -> !torch.list<!torch.float> {
%0 = torch.prim.ListConstruct %arg0, %arg0 : (!torch.float, !torch.float) -> !torch.list<!torch.float>
return %0 : !torch.list<!torch.float>
}
```
which turns into:
```
builtin.func @forward(%arg0: f64) -> !iree.list<f64> {
%c1 = constant 1 : index
%c0 = constant 0 : index
%c2 = constant 2 : index
%0 = iree.list.create %c2 : !iree.list<f64>
iree.list.set %0[%c0], %arg0 : !iree.list<f64>, f64
iree.list.set %0[%c1], %arg0 : !iree.list<f64>, f64
return %0 : !iree.list<f64>
}
```
As part of doing this, I realized that it was time to formalize the IR
form that we reach right before running TorchTo{Linalg,Std,...}. We now
call it the "Torch backend contract". We then lower the "Torch backend
contract" to the "npcomp backend contract", which involves the new
TorchConversion (`torch_c`) dialect, which holds ops that need to
operate on both the npcomp backend types (e.g. builtin tensors, i1, IREE
list, etc.) and the `!torch` types.
This made more sense, as I realized that if I didn't factor out
`torch_c` then the Torch dialect would have a dependency on IREE
dialect (we previously didn't notice this was an issue because we only
depended on `builtin` types), which seemed wrong to me.
Recommended review order:
- TorchToIREE.cpp / `TorchToIREE/basic.mlir`
- Look at the new structure of createTorchScriptToNpcompBackendPipeline.
It now lives in TorchConversion/Transforms/Passes.cpp and cleanly
calls into `Torch::createTorchScriptToTorchBackendPipeline` for the
frontend lowering to the Torch backend contract.
- Mechanical change extracting
`torch_c.{to,from}_{i1,i64,f64,builtin_tensor,iree_list}` into a new
TorchConversion dialect, and a few passes specific to the lowering
from the Torch backend contract to the npcomp backend contract.
- Minor fixes to TorchToLinalg.cpp to use unconverted operands (now that
we convert lists as part of operand materialization, we need to use
the original operands). Also added test for AtenMaxPool2dOp and fixed
m_TorchConstantIntList.
- TmpDeleteDeadIREELists pass. Temporary pass for deleting dead IREE lists that
are created as part of operand materialization for conv/max pool/avg pool ops
in TorchToLinalg.
With the following changes the compilation can continue until
RefineTypes pass:
- Add operators without ODS into `torch_ods_gen.py`
- Add some new optional and list types in `TorchTypes.td`
- Add some folders for aten int type comparator ops
- Modify GlobalizeObjectGraph.cpp. For global slots that's not used,
dont check if an aliased value is stored in more than one of global
slots. This can work around a failure where the same tensor is stored
in multiple "version" slots which are not used.
This includes the following changes to import MT model into MLIR. There
are still a lot of work to for actual compilation.
- Add `torch.dict<>`, `torch.any`, `torch.number` types
- Add `torch.prim.DictConstruct` op
- Fix `torch.prim.TupleConstruct` op assembly format to include resulting types
This takes the example from torchscript_resnet18_e2e.py and puts it into
a slightly cleaned up notebook form.
It's still a little rough around the edges. Areas for improvement:
- Installation / setup.
- API usability.
Also,
- Add `npcomp-backend-to-iree-frontend-pipeline` since we will be adding
more stuff there.
- Slight cleanups.
- torch.aten.flatten.using_ints to linalg lowering
- torch.aten.max_pool2d to linalg lowering
- Support torch.aten.conv2d for more flexible dilation and strides values
These were legacy concepts that are now superceded by direct Torch to
linalg-on-tensors lowering. These were based on some very early thinking
related to the layering of frontends vs codegen, which is now obsolete
because:
- We expected a lot more centralization at the frontend (TCF) level. It
turns out that frontend needs really vary a lot, and there is no grand
unifying TCF dialect plausible. The additional layer isn't worth it.
- Linalg-on-tensors obsoletes the primary need for TCP. There are still
a few things not representable with linalg-on-tensors, but the support
is growing and the whole "not included in linalg-on-tensors" direction
needs to be rethought. Our TCP dialect didn't cover any of the
actually important things in this space (such as sort, FFT, top-k,
etc.).
See historical [slides](https://drive.google.com/file/d/1iljcpTQ5NPaMfGpoPDFml1XkYxjK_6A4/view) / [recording](https://drive.google.com/file/d/1jSPa8TwPKUt0WuLquGc8OgSUVYJHMvWZ/view)
for more details on the origin story here.
Their presence was confusing users too
[bug](https://github.com/llvm/mlir-npcomp/issues/248).
Also,
- Trim down npcomp-run-mlir testing. It was testing TCF to TCP
lowering for the most part. The essential stuff is retained and
rephrased with linalg-on-tensors. (we should probably rename it
"refback-run" or something, as it is just a way to invoke RefBackend)
- test/Python/Backend/RefJIT/simple_invoke_numpy.py is XFAIL'ed. Our
"anti-framework" direction seems to be the likely future path.
* Change aligned_alloc -> malloc. It can fail (and does on MacOS) and is a bit over-aggressive optimization for a reference backend.
* Fixed a fragile test that prints -0.0 on MacOS.
* Fail the test (not the framework) on failure to trace (Torch on MacOS is missing features).
* Fix .so -> .dylib for compiler runtime.
* Added additional *ToLLVM conversion patterns (they were disaggregated from standard).
* Misc renames.
* Spelling change on ConvNCHW op, and it now expects strides and dilations attributes.
- Build adjustments for `.cpp.inc` dialect files.
- Renaming of `memref.dim` to `tensor.dim` for tensor case.
Minor changes:
- Renaming of `mlir::linalg::ReassociationIndices` to
`mlir::ReassociationIndices`.
- Adjust command line option parsing in npcomp-run-mlir.
We plumb through e2e a fair number of interesting cases:
- unary, binary, ternary elementwise ops
- ops like `torch.aten.add.Tensor` that also take a scalar parameter
- static size-1 broadcasting
We allow the static size-1 broadcasting case, but emit a runtime error
in the case of dynamic size-1 broadcasting. This seems like a sweet spot
subset of things that can be lowered directly to linalg, while not being
overly constraining to users. This is consistent with what IREE is doing
for CHLO->Linalg lowering as well
([code](50bf7a87e4/iree/compiler/InputConversion/MHLO/BroadcastingToLinalgPatterns.cpp (L1))).
To test the static size-1 case, we added support for the
`torch.aten.unsqueeze` op and lowering for it through
`linalg.tensor_expand_shape`. This involved a generalization of
`MaximizeValueSemantics` able to handle it (the solution there also
works for `torch.aten.flatten.using_ints` which we need for ResNet
anyway)
Also, a few minor additional changes:
- Add `VerifyInvariantsBeforeBackendLowering` pass, which catches a
large class of errors before we get to backend lowering (now that we
are doing dialect conversion, the errors are way nicer if we just emit
them up front rather than in the guts of a random pattern).
- Minor change to RefBackend to allow `linalg.tensor_expand_shape`.
Recommended review order:
- e2e tests in elementwise.py
- `ConvertElementwiseOp` in TorchToLinalg.cpp + elementwise.mlir test
- `ConvertAtenUnsqueezeOp` in TorchToLinalg.cpp + unsqueeze.mlir test
- RefineTypes.cpp + tests
- MaximizeValueSemantics changes + test
- VerifyInvariantsBeforeBackendLowering pass + test
This adds a pattern to MaximizeValueSemantics which does a simple
abstract interpretation within a block, which handles simple cases of
`torch.overwrite_tensor`, enough to remove all the unnecessary uses of
non-value tensors in ResNet right now.
Before/after IR:
[gist](https://gist.github.com/silvasean/a3e1ef625b19dfc63579f73cd3b543b6)
Also,
- Split `torch.copy.tensor` into `torch.copy.to_tensor` and
`torch.copy.to_vtensor` which convert between value and non-value
semantic tensors. This is a much cleaner factorization as they have
very separate use cases and properties (e.g. different side effects)
- Remove the various canonicalization patterns they had, which were
confusing because they resulted in limited forms of maximizing value
semantics throughout the pipeline. We should structure our compilation
pipeline such that only MaximizeValueSemantics should be maximizing
value semantics.
- Adjust pass pipeline to only run MaximizeValueSemantics once.
- Make OverwriteTensorOp `$value` always be a value tensor and
`$overwritten` be a non-value tensor.
1. Added a simplified version of torch.aten.batch_norm which only handles
inference and assumes the weight, bias, running_mean, running_var are not
None.
2. Removed the primitive types check in verifyLinalgCompatibleTypes check
since now we have proper type converter to handle torch types conversion.
The checks for RankedTensorType is kept because the type converter
doesn't guarantee the converted builtin tensor type is ranked. A
separate verification pass to verify the invariant expected by later
passes will need to be added before those can be removed as well.
This op is much better behaved than the `torch.tensor.literal` op
(which is the new name of the `torch.tensor` op). In particular
`torch.tensor.literal`:
- always has a maximally refined type.
- always has value semantics.
- can be constant folded / CSE'd.
ReduceOpVariants is changed to perform the transformation from
`torch.tensor.literal` to `torch.vtensor.literal` (which in general
involves static information casts and copies.
This new op also allowed tightening up `torch.tensor.literal` to only
accept NonValueTensorType (instead of any tensor type).
This new ".literal" name is more descriptive. It was getting too
confusing seeing an op called just `torch.tensor` (we originally called
it that because that's the name of the similar function in the Torch
Python API, but it just doesn't fit here).
This removes the dependence of the `torch` dialect on the low-level
builtin types.
Now the `torch` dialect is a standalone layer, suitable for targeting
from higher-level Python abstractions without any premature lowering to
primitive types.
This replaces the ad-hoc use of `i64` throughout the Torch layer, and
helps to keep it crystal clear the distinction between `!torch.int`
(which is modeling the Python `int` type) and the various types that
serve as dtypes of tensors, which are a totally different type universe.
Changes:
- `!torch.int` type and C bindings.
- Change `torch.constant.int` parser to not need the `: i64` at the end.
- `m_TorchConstantInt` matcher to aid with matching constants.
- BackendTypeConversion changes for `!torch.int` -> `i64` type
conversion.
- Refactor finalizing patterns in FinalizingBackendTypeConversionPass
(they were getting very repetitive).
- Mechanical rewriting of `!torch.int` to `i64` in all the tests, and
`AnyTorchIntType` to `Torch_IntType` in the `.td` files.
This fixes a "regression" on ResNet where we weren't folding away all
the control flow. For now, our policy is to "optimize hard enough" to
make that control flow go away, because we don't yet have a way to lower
to the backend the stuff guarded by the control flow (RaiseException,
string operations, etc.).
It remains to be seen how much optimization we decide to do at this
level in the fullness of time -- the torch op set is not particularly
well-designed (at least not idiomatically for MLIR) for general
optimization. Ideally, with really good backend support for various
features, all the heavy optimization will happen at that layer on `std`
ops and `scf` control flow. But I have a suspicion we might end up
needing more optimization earlier in the pipeline.
This removes the use of `scf.if`, which required laundering back and
forth between `i1` and `!torch.bool` in the frontend. We will eventually
lower this op to `scf.if`, but this results in a cleaner IR and layering
at the frontend.
This finishes removing the dependence on the basicpy dialect!
Changes:
- Add `!torch.bool` type and replace use of `!basicpy.BoolType` in
Torch-related code.
- Rename BuiltinTensorize to BackendTypeConversion since now it handles
bool conversions (and, when we add !torch.int and !torch.float, it
will handle those as well), and generalize the related utilities (I
also moved them to Torch/Transforms since they aren't really part of
Torch/IR).
- Add `torch.to_i1` and `torch.from_i1` ops for materializations
- [cleanup] Reorganize `torch.constant.*` ops in TorchOps.td
- Remove dependency of `torch` dialect on `basicpy` dialect and also
`std` dialect. For `std`, we use some call related ops, but the
`torch` dialect itself never produces them (we have passes that do
though).
This is fairly mechanical. Recommended review order:
- New stuff in Torch/IR
- New BuiltinTypeConversion files.
- Mechnical fixups elsewhere.
- Make consistent with MLIR Core
- Use `//` or `///` comments.
- Use `bool` type for booleans
- No duplicated comments in .cpp files
- Split types into separate files `{Basicpy,Numpy,Torch}Types.h`
- Add dialect prefix consistently to C API symbols. We have lots of
similarly named types (e.g. "list" type in basicpy and torch).
- Add `torch.constant.none` op to construct it (naming is chosen to be
analogous to Torch's representation of a prim::Constant with
NoneType, rather than using the "singleton" terminology of Basicpy).
This removes our reliance on the numpy dialect and avoids our off-label
use of the builtin tnesor type for modeling unknown dtypes. The
`!torch.vtensor` (`ValueTensorType`) type is a value-semantic tensor.
The `!torch.tensor` (`NonValueTensorType`) type is a non-value-semantic
tensor. The new types look as follows syntactically:
```
// Least-static-information, non-value-semantic tensor.
!torch.tensor
// Explicit form of least-static-information variant.
!torch.tensor<*,unk>
// Least-static-information, value-semantic tensor.
!torch.vtensor
// Explicit form of least-static-information variant.
!torch.vtensor<*,unk>
// Fixed-set of allowable element types, with first-class support for
// Torch's frontend signedness semantics.
!torch.tensor<*,si32>
// First-class support for unknown dtypes.
!torch.tensor<[?,?,?],unk>
// Standard MLIR representation of `?` for unknown dimensions.
!torch.tensor<[?,2,?,4],unk>
// Statically shaped / dtyped example.
!torch.vtensor<[1,2,3,4],f32>
```
This required fairly significant changes throughout the compiler, but
overall it is a big cleanup. We now have a much clearer layering of "the
Torch frontend lowering" vs "lowering to std + linalg + etc.".
At the C++ level, there is `ValueTensorType`, `NonValueTensorType`.
We also have a helper `BaseTensorType` (kind of like ShapedType) which
interoperates with those two.
Included changes:
- New `torch.tensor(dense<0.0> : tensor<5xf32>) : !torch.tensor` op for
creating torch tensor literals in the frontend.
- Consistently use signedness for the types (except i1 which I didn't
touch -- we need to sort out the situation with !basicpy.BoolType
there anyway so will be attending to that soon)
- Frontend can annotate whether an argument to the function has value
semantics. We currently require this, as our backend contract does not
currently allow us to even model the non-value-semantic case. Before,
the value-semantic assumption was randomly injected in the middle of
the pass pipeline.
- Move ArrayToTensor (now called MaximizeValueSemantics) and
RefinePublicReturn passes to torch dialect.
- The TorchToStd and TorchToLinalg passes are now type conversions from
`!torch.vtensor` to `tensor` and use the dialect conversion infra.
The overall conversion pipeline is set up following the best practices
of the "Type Conversions the Not-So-Hard Way" talk. This required
introducing `torch-func-builtin-tensorize` and
`torch-finalizing-builtin-tensorize` passes analogous to the upstream
bufferization passes with the corresponding names (mostly just
copypasta from there).
- Misc Torch-level canonicalizations -- we now cleanly layer the
lowering to std later in the pipeline, so we are gradually lessening
our reliance on random std constant folding before we get to that
point.
Recommended review order:
- New types in TorchTypes.td/TorchTypes.h/TorchDialect.cpp
- New ops in TorchOps.td / TorchOps.cpp
- Less important / more mechanical stuff
- Frontend changes.
- Pass changes/additions in `Torch/Transforms` and `Conversion/`
This now gives [much nicer output](https://gist.github.com/silvasean/f048e0f37b04542dae6469b86802bb3e).
Embarrassingly, we previously couldn't even report failures for two
different tests, and weren't able to report on compilation failures
(besides just crashing).
This is enough to import the program and get it through the compilation
pipeline. It of course fails at the VerifyBackendContract pass since
there is a lot missing, but the final IR for a simple quantized MLP is
looking pretty decent already:
[IR](https://gist.github.com/silvasean/f76bccd76e9b193d396cfb2f9a11f54d)
Main changes:
- Add support for importing torch quantized tensors, including
`torch.per_tensor_affine.create` op and `!torch.qint8` element type.
- Add support for importing `LinearPackedParamsBase` (basically a weight
+ optional bias, but requires `torch.linear_params.create` op +
`!torch.LinearParams` type to model it). This was less painful than I
expected, as it has the necessary methods to opaquely unpack itself. I
factored things so it should be easy to extend to other custom classes
like `ConvPackedParamsBase`.
- Add minimal boilerplate for importing `quantized::*` ops, with
`quantized::linear` being a motivating example.
- Add e2e test with simple quantized MLP (courtesy of @phoenix-meadowlark).
This is somewhat of an abuse of `!numpy.ndarray` / `tensor`, as
really the proper semantics of `!torch.qint8` dtype on a Torch tensor is
"check the quantizer object of the tensor for side data (scale/offset,
possibly per-channel) that defines the full semantics of the tensor". We
don't have any such notion of "side data" for `!numpy.ndarray` /
`tensor`, let alone anything that would have the associated behavior of
keying off the dtype to determine if the side data is present.
This will be fixed by a proper `!torch.tensor` type.
This is a really major and invasive restructuring of the way we get
torch operators (`torch::jit::Operator` / `c10::OperatorHandle`) into
MLIR. Please forgive the challenging review, but due to the sheer
invasiveness, it wasn't really practical do do it in sane smaller
pieces.
This fully replaces everything that was already working on the
TorchScript path (actually, more -- we added tanh support to
TorchToLinalg in order to delete the older code paths). Additionally,
I've kept the lights on for the acap path too, including what little e2e
stuff was working before (for expediency I made a few tiny compromises
along the way that will be easy to undo when we give that path proper
attention).
Overview of the new design:
- The torch operator `somens::someunqualname.someoverloadname` is
imported as `torch.somens.someunqualname.someoverloadname` (skip the
last dotted part if the overload name is empty), OR, if we don't have
such an op registered, it is imported as
`torch.operator "somens.someunqualname.someoverloadname" (...) : ...`.
- The addition of the "overload name" is a critical element here, as
the `(ns,unqual,overload)` triple is unique, which solves a lot of
problems we were having.
- This involves having separate MLIR ops for the `trailing_` and
`.out` variants and all the different overloads. This seemed
necessary, because the set of overloads is so wild and varied and
unstructured. The previous design was leaning into some underlying
structure that just isn't there -- the default situation is
the "random overload that we want to manage on the MLIR side",
rather than that being an exception. E.g. `aten::ne` (not-equal)
has 21 overloads, only 4 of which are c10 dispatcher ops see
[gist](https://gist.github.com/silvasean/190ba918c550c956260e21254e1b8aa1),
and the "out" variant is really called `.Tensor_out` instead of
`.out` as it frequently is for other ops.
- Rationale for all being in `torch` namespace: the set of operators
are so varied and unstructured that "dialect per namespace"
doesn't result in anything resembling the typical MLIR dialect
boundary expectations. We could maybe draw the boundary at
dispatcher ops vs non-dispatcher ops, but that doesn't seem to
really result in very much useful structure at this point in time.
- Note: within the torch operator registry, we effectively have a
mini-basicpy subdialect (already type-resolved), which is reasonably
structured.
- The existing Torch op interfaces are also removed -- now that we
track the overload name, we can losslessly find the original
operator.
- Instead of `ATenRecognizeKernelsPass`, we now have a
`ReduceOpVariantsPass` that keys off certain traits (and perhaps
eventually interfaces) to reduce variants of ops to a smaller set,
ideally operating on immutable tensors and using surrounding ops to
model the mutability/aliasing aspects.
- Note: `torch.ns.unqual.overload` ops allow both immutable and
mutable tensors (unlike the previous hard distinction in the common
case). This is a premonition for a future change that will introduce a
bona fide `!torch.tensor` type that will clean up a bunch of stuff.
- `TorchToLinalg` / `TorchToStd` supercede the existing
"ATen->TCF->TCP->Linalg" path.
- The new `torch_ods_gen.py` supercedes `torch_signature_ods_gen.py`.
It should look somewhat familiar, but the benefit of hindsight has
allowed a lot of simplifications.
The overall trend seems to be to make the `torch` dialect a nice layer
independent of anything else. It feels like as a natural result of
various future changes we will be removing the reliance on basicpy+numpy
dialects and have a nice self-contained type system too that properly
models the TorchScript type system (including proper subtyping,
mutable/immutable tensors, optional dtype, etc.).
Recommended review order:
- Start at some of the new import IR, e.g. in
`frontends/pytorch/test/node_import/prim.py`,
`frontends/pytorch/test/acap_export/test_export_add3.py`, and other
tests.
- `frontends/pytorch/python/torch_mlir_utils/codegen/torch_ods_gen.py`
and associated generated files:
- `include/npcomp/Dialect/Torch/IR/GeneratedAtenOps.td`
- `include/npcomp/Dialect/Torch/IR/GeneratedPrimOps.td`
- Inspect `ReduceOpVariants.cpp` / `reduce-op-variants.mlir` and the new
traits in `include/npcomp/Dialect/Torch/IR/TorchTraits.h`
- Various code changes in the import path in
`frontends/pytorch/csrc/builder`. Probably most interesting is the new
code in `torch_to_mlir_utils.cpp` that has the logic to create the
`torch.operator` ops or `torch.ns.unqual.overload` ops.
This is the [new ResNet IR](https://gist.github.com/silvasean/5407aafb710d07612b7b5b92eabecebe),
just to be able to look at a substantial sample of IR in the new style.
- aten::relu_, aten::max_pool2d, aten::adaptive_avg_pool2d, aten::batch_norm, aten::conv2d
No aten-to-linalg conversion for the latter ones, as they are fairly
substantial. At this point, I'm trying to get shape inference and stuff
working for them and the IR cleaned up.
This trait lets us model the semantics of various aten/torch/numpy ops
that are insensitive to type refinements. This replaces
hardcoded/inconsistent checks for this property.
To show usage of this new trait, we fix up some old uses, and improve
RefineTypes to be smarter about rewriting with this trait.
Interestingly, TorchScript has its own op (`torch::jit::Operator`)
registry separate from the dispatcher (it is a superset of the
dispatcher).
This is where the "prim" ops and some "aten" ops (that should probably
be renamed to "prim") live. In particular, `aten::__is__` is in that
latter category of "aten but really prim". This registry is also the
source of truth for what the TorchScript interpreter calls into when it
executes.
The bulk of the "not part of the dispatcher" ops live in
09feb5f579/torch/csrc/jit/runtime/register_prim_ops.cpp (L82)
And the registry itself lives in:
09feb5f579/torch/csrc/jit/runtime/operator.cpp (L196)
This fold further reduces the IR of ResNet by folding away some
more not-taken branches. These not-taken branches in ResNet require
first-class handling of the list type which we don't yet have on any
backend.
This is the start of a push to getting ResNet running.
This involves throwing in the towel on an O0 pipelinie for now. See note
in the code. We keep an options struct with `optimize` flag, but it
default to true for now.
This removes the need for defining all of the custom propagation logic,
and also adds support for propagating value knowledge across branches,
through regions, and across calls.
These tests pass on the reference backend.
- Add aten.linear op + shape xfer function + ATen->Linalg lowering.
- Note: this needs to be more automated, and needs to cover more cases.
- Current not implemented caveats:
- size-1 broadcasting for bias vector (either static-size-1 or ? case)
- higher-rank aten.linear ops (not produced by torch.nn.Linear though)
- type promotion (still don't even know the exact rules here)
- Add folder for torch.derefine op. Now the inliner can clean it up as
it inlines. (call boundaries are a main place we need to insert
torch.derefine) This is brittle -- the other important case is control
flow which will need to be handled via an extension to
RefineTypes.cpp (as will more robust call handling). River has an
in-flight patch to update it to the new dataflow framework so I didn't
want to do anything intrusive here.
- Also adjust torch.derefine syntax to use the keyword `to` instead of
`->`, as most type-only, cast-like ops do.
This inlines global slots if possible. This allows them to participate
in folding, canonicalization, shape inference, etc.
Example use cases:
- inlining weights and biases that are readonly during inference
- inlining the "training" bool to allow stuff to fold away
For training use cases (especially internal training loop), we will need
something smarter to get good performance. That would look like an "SSA
formation" which promotes the global slots to tensors in the program,
flushing them back to the slots at the minimal number of necessary
places. We might want to let backends do that transformation though.
This also interacts with shape inference (type bounds on the slots to
even lower them to backends in the first place).
- Move frontend lowering pipelines to c++ (this helps with reproducing
failures in npcomp-opt)
- Add debugging printouts when compilation fails on RefBackendTestConfig
The experience now when a test fails during MLIR lowering is now like this:
```
NPCOMP TorchScript Object Graph IR -> NPCOMP Backend IR lowering failed with the following diagnostics:
failed to legalize operation 'torch.global_slot'
Module does not conform to npcomp's backend contract. See dialect conversion legality information above.
Error can be reproduced with:
$ npcomp-opt -torchscript-to-npcomp-backend-pipeline /tmp/ResNet18Module.mlir
```
And when TorchScript->MLIR import fails it looks like this:
```
PyTorch TorchScript module -> NPCOMP Object Graph IR import failed with the following diagnostics:
unhandled prim operation: %18 : int = prim::min(%17) # /usr/local/google/home/silvasean/.local/lib/python3.9/site-packages/torch/nn/functional.py:4532:4
```
Also,
- Add `--filter=<regex>` to e2e test harness to filter tests.
- Add a few prim ops that were needed to import ResNet18
- Fix torch.prim.Loop.condition assemblyFormat (it previously would not
round-trip in the case of no loop-carried variables)
As described in the code comment:
```
When we import TorchScript IR, we import their entire "compilation unit",
which can contain numerous functions unrelated to the current program,
which breaks torch-globalization-pipeline; for example, there can be
random functions referencing types that haven't been imported
as part of the root `torch.nn.Module` we imported. Those will
be unreferenced private functions which symbol-dce will clean up nicely.
```
This situation is really easy to hit in jupyter notebooks, where the
same cell is evaluated multiple times. That results in the same
class name (at the Python level, e.g. class `Foo` in the top-level
main module). Internally to PyTorch, it handles this situation by
mangling in a unique number to the names of ClassType's and such. When
we import the new ClassType's, we see not just the new
torch::jit::Function's in the CompilationUnit, but, also all the old
ones, which reference ClassType's that are not reachable from the
`torch.nn.Module` that we imported.
Note: there is no way to avoid importing the whole CompilationUnit
(including these old remnants) without doing a fairly complicated call
graph reachability analysis of which functions are reachable from the
methods of the ClassType's we imported. It turns out that once we are
inside MLIR, we model visibility correctly so that `symbol-dce`
"Just Works" for this use case. That is to say, this is not a quick
hack, but rather seems like a totally palatable long-term solution.
This pass verifies that a given module satisfies the contract that we
have for backends. This is phrased as an "allowlist", because we want to
keep this interface tight. Also, this gives much better diagnostics than
a backend randomly crashing or failing to compile would (though they
could still be improved).
This was especially painful because if we had
`tensor<?x!numpy.any_dtype>` slip through, at some point RefBackend
would convert it to a memref type and trip the "verify type invariants"
assertion which gives no location or anything and crashed the process,
which was very unpleasant.
We implement this with the dialect conversion framework, which works
reasonably well and was quick to put together and familiar, but is still
very "op oriented". We probably want to make this hand-rolled
eventually, especially the error reporting (the most useful kind of
error for a dialect conversion user is not necessarily the best for this
use case). Also, in production, these error will go to users, and need
to be surfaced carefully such as "the compiler needs a type annotation
on this function parameter" which in general requires some special
analysis, wordsmithing, and overall awareness of the e2e use case (such
as how much we can lean into certain source locations) to provide a
meaningful user-level diagnostic.
Also, add `inline` to the current frontend lowering pass pipeline to
allow slightly more complicated programs that otherwise would fail on
shape inference.
This is our first op with error semantics, and stresses the system.
There are a few design notes of special interest:
- RefineTypes.cpp's note about shape inference in the presence of code
that dynamically produces and error, and it is provable statically.
- ATenToLinalg.cpp's notes about future automation of the ATen->linalg
path.
- The notes in Passes.td about using low-tech `std.assert` ops instead
of `shape.assuming`.
Note: Doesn't work on IREE yet due to the `std.assert` op (needs to be
lowered to `vm.fail` on the IREE side).
Recommended review order:
- Changes in frontends/pytorch/examples/
- Changes in python/npcomp/compiler/pytorch/backend/
- Boilerplate for the `npcomp-iree-backend-lower-linkage` pass.
This change separates out a
`npcomp.compiler.pytorch.backend.frontend_lowering` module that does the
common lowering for all backends. The individual compiler backends
`npcomp.compiler.pytorch.backend.{refjit,iree}` now accept a loosely
defined "TCP + scalar code" IR mix that will be formalized in the
future as the interface to codegen backends.
This also required adding a small pass
`npcomp-iree-backend-lower-linkage` which adds `iree.module.export` onto
functions, and layering that into the frontend flow. The pass doesn't
require a C++-level dependency on IREE, which is nice for now. TBD how
we are going to handle lists (we hope we can get away with sneakerneting
some td files and relying on loose IR compatibility).
Running through IREE requires the ability to import `iree.compiler` and
`iree.runtime`, which can be obtained as follows:
```
python3 -m pip install iree-compiler-snapshot iree-runtime-snapshot -f https://github.com/google/iree/releases/tag/snapshot-20210406.200
PYTHONPATH="${PYTHONPATH}:${MY_IREE_BUILD}/bindings/python/"
```
This patch makes it painfully clear that we don't have any e2e testing
harness to really plug into, and also don't have a usable Python API to
our compiler stack (something usable in a jupyter notebook).
That will be addressed in subsequent commits. We've been flying by the
seat of our pants with this `examples` directory that isn't subject to
any kind of testing or real usability concerns.
This pass allows shape information to be propagated to return types,
which is nontrivial and cannot be cleanly put anywhere else as it
changes the public ABI, which is a concern that we want to keep
concentrated in one place.
Currently implemented as a simple intraprocedural dataflow analysis over
a standard ShapedType lattice (hasRank, sizes, and elementType).
It currently hardcodes a few key pieces of information:
- shape transfer functions
- whether it is legal to update the operand type of an op
This needs to be made pluggable obviously and the core propagation logic
moved somewhere agnostic.
The current implementation is just sufficient to do a unary aten.tanh
from the e2e spike, and just applies some local rewrite patterns. I've
sketched out the more full explanation of where this pass eventually
need to go in the pass docs.
Adding this required adding `numpy.tensor_static_info_cast`, which is
the tensor analog of `numpy.static_info_cast`. This op encapsulates the
same numpy-specific "no runtime code" casting semantics, in particular
the interpretation of `!numpy.any_dtype`. The
`numpy.tensor_static_info_cast` I see in practice now are "information
erasing" and will be removed by a later pass that exploits the fact that
aten ops are agnostic to the static info in the operand types (so
substituting a type with more static info is fine).
Side note: we *need* to do dtype and rank inference before aten->tcf
(which will eventually mostly be aten->linalg+guards), because each aten
op is idiosyncratically overloaded based on dtype and rank. Without
copying that idiosyncratic overloading into lower layers (layering
violation), we cannot really lower it to anything until we do that.
This pass incorporates torch.type_bound info and also removes NoneType
returns (eventually it will rewrite tuple types too, but can't yet
because !basicpy.TupleType doesn't track element types).
Recommend looking at adjust-calling-conventions.mlir first to see what
it is doing, and holding your nose for the implementation of the pass.
I decided to implement this with the conversion framework, because it
gives us *some* goodies for type conversion -- mainly avoiding large
amounts of tricky RAUW dances. Unfortunately, the conversion framework
isn't a perfect fit for a couple reasons:
- the incorporation of torch.type_bound is a context-sensitive rewrite
(requires looking at the arg attr, not just the type).
- NoneType conversion is 1->0, which requires some special handling
- (not implemented yet) 1->N tuple type conversions require special
handling.
It's a little bit scary, but on balance doing it the other way would
have its own downsides.
These allow users to annotate a known "type bound" on the argument,
which can seed shape/dtype inference. We don't rewrite the function
types as part of the import process (it will happen in a
yet-to-be-written pass) because:
1. We would need to interprocedurally rewrite all calls to keep the IR
consistent. Currently, we have a place after GlobalizeObjectGraph but
before we convert to tensors where this is convenient to do. Ideally,
we would do this on the object graph representation.
1. We don't necessarily know that adjusting the function type is a legal
calling convention change. The pass will have blessed knowledge (by
the pass pipeline author) that adjusting the argument type based on
the type bound is safe (which it frequently is).
2. Note that in principle, a type bound could be a fairly general thing
(such as maximum sizes of dimensions, unions of multiple concrete
types, etc.). The pass will in principle have logic to interpret the
type bounds and to determine a suitable "best" (and legal) argument
type.
- renames of OwningRewritePatternList -> RewritePatternSet
- also `insert` to `add`
- RewritePatternSet holds a context now
- memref dialect split from std
* Adds f32 scalar argument support across the ABI boundary.
* Adds support for passing input type / shape information
across the ABI boundary
* Adds support for parsing / creating input FloatAttr's in
`npcomp-run-mlir`
We already had the `promoteTrailingOutTensor` flag, but weren't using
it. A inplaceVariantKernelName flag needed to be added.
This change is a little dissatisfying, as the conversions done by the
RecognizeKernelsPass are currently non-orthogonal. In particular,
`kDropResultAndAliasArg0` probably won't work as intended if mixed with
these (we probably need to promote kDropResultAndAliasArg0 to not be an
arg-level thing anyway, as we have done with promoteTrailingOutTensor).
This involved adding a new op `numpy.overwrite_array`.
```
numpy.overwrite_array %arg2 overwrites %arg0 : tensor<2x3xf32>, !numpy.ndarray<[2,3]:f32>
```
This models the destructive update behavior. Note that in the above op,
we cannot simply RAUW %arg0 with a suitably conveted %arg2 (for example,
%arg0 might have uses that are not dominated by %arg2, or might have an
alias relation with some other array in the program). In general, we
need a pass analogous to "SSA-formation" which knows how to see through
these to uncover an underlying tensor program.
Also, add tanh_out_e2e.py/div_inplace_e2e.py and fix some bitrot in
refjit.py which is my running example I'm trying to get working.
* Import ATen conv2d conversion and test
This is a first attempt at expanding ATen-to-TCF conversion for the
conv2d operator. Eventually, this will come in use when lowering a
high-level conv-based model.
This happens in practice with e.g. ResNet from torchvision (multiple
instances of the same BatchNorm class).
The key observation is that for this program, and the expected set of
programs, we can convert the program to the same globalized form with a
bit more static analysis and effort to suitably monomorphize the
program. Though what we are doing here is fairly annoying to implement,
it saves any nontrivial later pass from having to do similar analyses
(or worse). E.g. shape inference would need to be object-graph aware,
mutation/lifetime analyses would have to be aware, etc. Additionally, it
would make us front-load what it means to have a !torch.nn.Module type
on an ABI boundary, which we are just not ready to handle.
I'm really, really hoping that in practice we can get away with
this, otherwise it's going to be really rough designing a representation
(and implementing everything to back it) that is convenient to transform
and gracefully scales from full object graph (in the most dynamic case)
down to a fixed set of global slots like we have here (in the most
static case, which we presume a lot of practical programs fall into).
This also involved introducing a
`torch-prepare-for-globalize-object-graph` pass that does a minimal set of
lowerings to simplify the IR into a more orthogonal and analyzable form,
and a `torch-globalize-pipeline` helper.
Recommended review order:
- updated documentation in Passes.td
- new tests in `globalize-object-graph-multiple-instances*.mlir`
- implementation of GlobalizeObjectGraph.cpp
- PrepareForGlobalizeObjectGraph.cpp + prepare-for-globalize-object-graph.mlir
- misc stuff like torch-globalize-pipeline pipeline definition.
With this, we can import, globalize, and inline resnet18 from
torchvision:
https://gist.github.com/silvasean/821586afc19b67d9fb72030b2e0adeb8
In terms of IR structure, TorchScript allows types to vary in many
circumstances where MLIR requires pointer-identical types. In particular,
it is valid to pass any subtype in place of a type. For example, if an
`Optional[int]` is required somewhere in the IR, it is legal to pass a
value of just `int` (but not the other way around; see
`torch.prim.unchecked_cast`). In effect, every *use* can have a different
type.
We introduce a new op `torch.derefine` that models that impedance
mismatch. This op allows casting a value from one type to a type that it
is a subtype of to model this behavior.
Recommended review order:
- TorchOps.td for new torch.derefine (and updated docs for
`torch.prim.unchecked_cast`)
- new test code in if.py, loop.py, function-derefine.py
- new code in node_importer.cpp for handling derefinement insertion
- function_importer.cpp and utils changes in torch_to_mlir_utils.cpp
Properly handling derefinement on function boundaries required
relayering the code so that graph_importer.cpp/.h is now
function_importer.cpp/.h because only the `torch::jit::Function`
(actually the `c10::FunctionSchema` it holds) knows the derefined types that are
actually needed at the boundary (see `function-derefine.py` for a test).
Annoyingly, this churns all the functions which are now prefixed with
`__torch__.` but that is more correct anyway (that is their linkage name
in the `torch::jit::CompilationUnit`; the previous `mb.import_function`
was actually buggy in the case of functions calling each other as it
would reference their unqualified name).
With this change, we can import `resnet18` from `torchvision` :)
IR: https://gist.github.com/silvasean/6426a5272d8a6c7caae533fce05ab704
This primarily unlocks proper handling of free functions (that is,
functions that are not methods of any torch.nn.Module).
Recommended review order:
- `ivalue_importer.cpp` + `ivalue_import/functions*.py`
- `GlobalizeObjectGraph.cpp` + test case
- misc other stuff
The `torch::jit::CompilationUnit` is basically a backing store or
"context" holding all the possible functions in the program. The
previous code was not explicitly accessing this data structure, since it
just imported the `torch::jit::Function`'s that it saw attached to
methods.
Subtly, any time a TorchScript module called into a free function, the
free function gets incorporated into the torch::jit::CompilationUnit,
but doesn't show up anywhere when dumping the module, except in the
curious pattern:
```
%5 : Function = prim::Constant[name="adaptive_avg_pool2d"]()
%6 : Tensor = prim::CallFunction(%5, %input.1, %4)
```
That is, calls are indirect calls, and are accessed via `prim::Constant`
materializing a function object. Even stranger, the `name` attribute here
doesn't really even tell the full story -- it doesn't correspond to
anything. It turns out that the c10::FunctionType itself actually holds
a pointer to the `torch::jit::Function` in the compilation unit
directly (so there is actually no indirection in prim::CallMethod,
because any two values of the same FunctionType call the same
function!). E.g. when converting the IR to bytecode, the "name" is
ignored [code link](1d6bd15790/torch/csrc/jit/runtime/interpreter.cpp (L937)).
We do import `prim::CallFunction` as a `std.call_indirect` though
because it's more braindead to do it that way (it gets canonicalized to
a direct call easily).
This is a much simpler representation than the ad-hoc initializer
function we had before. It is also less general, but given the rationale
in Passes.td it seems like the right tradeoff right now.
We can probably carry this representation for quite a while, and when we
can't, it likely means that TorchScript has fixed their object identity
bug and we probably need to just upgrade to a more general object graph
modeling (more general than GlobalizeObjectGraph).
In particular, we don't want to deal with defining and carrying around
this initializer function concept until we need it. For example, if we
want to constant-fold the global slots into uses, this is a much better
representation, and it plays better with symbol-dce (the initializer
function counts as a "use" of the symbol).
(the alternative would have been to write a pass that converts the
initializer function to this form when possible, but I realized that
lots of information had been lost which made that fairly annoying -- it
was all self-inflicted anyway, so best to just go to the source
(GlobalizeObjectGraph) before the information is lost)
Now symbol-dce works nicely (no more "training" bools)
```
pt_util ~/tmp/classifier.pt --import --exported-name forward \
| npcomp-opt -torch-globalize-object-graph -inline -symbol-dce
```
IR: https://gist.github.com/silvasean/8abe63d70d24e29d6db9170ccc8d512b
The first use case is to annotate certain program constructs as either
exported or private. In this commit we plumb it down to
GlobalizeObjectGraph which makes use of this information.
Recommended review order:
1. class_annotator.h/.cpp + `test/module_import/annotations/*`
- New abstractions to communicate with Python code and annotate.
2. IR changes in TorchOps.td
- Adding "private" attribute to various things.
3. ivalue_import.cpp changes
- Module + ClassAnnotator = annotated IR
4. GlobalizeObjectGraph.cpp + tests
- use new "private" attributes to create "private" IR.
- also, tweak some of the op deleting mechanics, which was triggering
some memory errors / assertions
With this, we can run the classifier through and inline it as follows:
```
frontends/pytorch/utils/pt_util.py --import --exported-name forward ~/tmp/classifier.pt \
| npcomp-opt -torch-globalize-object-graph -inline
```
IR: https://gist.github.com/silvasean/32dcad9f6270557f412094a77cecdd69
With this + manually setting private visibility on everything, a simple
classifier can be reduced to this IR, which is looking pretty lean and
mean:
https://gist.github.com/silvasean/19e7e2e21a61ff197aeac0dd864d188f
Also, include a utility script for importing `.pt` models.
```
pt_util.py --import classifier.pt | npcomp-opt -torch-globalize-object-graph
```
This required restructuring of how we model TorchScript on import. The
main difference is that now we split out a `torch.class_type` that holds
methods and declarations of the types of each slot. This is more
consistent with TorchScript (our previous representation was
"denormalized").
Recommended reading order:
1. check out the description of `torch.class_type` in `TorchOps.td` and
look at `test/Dialect/Torch/ops.mlir` and
`frontends/pytorch/test/module_import/` to familiarize with the new
representation.
- Just look at the new IR. The diff between the old names and new
names is confusing.
2. check out `test/Dialect/Torch/globalize-object-graph*.mlir`
and read along with the pass description in
`include/npcomp/Dialect/Torch/Transforms/Passes.td`
3. Read the code in `GlobalizeObjectGraph.cpp` and miscellaneous changes
in `ivalue_importer.cpp`, `TorchOps.cpp`, etc.
Sean Silva
Ramiro Leal-Cavazos
Prateek Gupta
Vivek Khandelwal
Gaurav Shukla
Yi Zhang
Prashant Kumar
Nirvedh
Anup Gangwar
Anup Gangwar
Suraj Sudhir
Prashant Kumar
Liam Fitzpatrick
stephenneuendorffer
dan
xndcn
Daniel Garvey
Suraj Sudhir
ds1231h
nodlabs
George Petterson
Wang Kangyu
Boian Petkantchin
Stella Laurenzo
River Riddle
Aaron J Arthurs
Bryce Arden
Aaron Arthurs
Bairen Yi