As per title. See also
[PR](https://github.com/llvm/torch-mlir/pull/3750) for
`torch.aten.mul.float_int`.
---------
Co-authored-by: zjgarvey <47986913+zjgarvey@users.noreply.github.com>
- Add Torch to TOSA legalization for the following ops:
+ aten.reflection_pad1d
+ aten.reflection_pad2d
+ aten.replication_pad2d
- Update xfail sets with new e2e results
- Add new LIT tests to basic.mlir
Change-Id: I1689d1778d8e472c3317aca1e2425ef8774a07fa
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
- Refactors more "onnx.ReduceXXX" patterns through helper function.
- Fixes bug with iterating unconditionally on `output_dim == 1` during
`dimList` inference.
This change results in passes for the following 11 models:
crossvit_15_240
crossvit_15_dagger_240
crossvit_15_dagger_408
crossvit_18_240
crossvit_18_dagger_240
crossvit_18_dagger_408
crossvit_9_240
crossvit_9_dagger_240
crossvit_base_240
crossvit_small_240
crossvit_tiny_240
---------
Co-authored-by: Vinayak Dev <104419489+vinayakdsci@users.noreply.github.com>
Essentially, as part of my earlier
[change](7f9f99c6f8)
, I didn't consider the `padding` value while erroring out for
unsupported `count_include_pad` during `torch-to-tosa` lowering for
AvgPool2d. The fix captured in this change addresses this. Please see
[issue](https://github.com/llvm/torch-mlir/issues/3862) for more details
on this.
Co-authored-by: Hanumanth Hanumantharayappa <hhanuman@ah-hhanuman-l.dhcp.mathworks.com>
- Fix aten.rsub.Scalar legalization with appropriate type casting
- Add legalization for aten.clamp.Tensor
- Resolve some unexpected test failures from PyTorch update by adding
legalization for the following ops:
+ aten.avg_pool1d
+ aten.max_pool1d
+ torch.prims.collapse
- Update xfail_sets with new e2e results
- Add new LIT tests to basic.mlir
Change-Id: I9762c7d36ca0b0f75ca68d0c71d7f5d5309a96ad
---------
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
- Add Torch to TOSA legalization for aten.as_strided op
- Update xfail_sets with the following:
+ New aten.as_strided results
+ Changes from this commit:
7f9f99c6f8
+ Failed tests from new PyTorch version update
- Add new LIT test to basic.mlir
Change-Id: I6f471ea116ca47f2bf9537b62950fce75a2c624f
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
Added lit tests since these scalar operations don't trace well through
the `fx_importer` route.
`XOR` and `NE` are equivalent binary operators, so `aten.ne.bool` is
lowered to `arith.xori`.
The existing TorchToTosa lowering logic for `torch.aten.avg_pool2d`
doesn't handle some unsupported properties well, leading to a silent
wrong answer(SWA) when we go through
`torch-backend-to-tosa-backend-pipeline.` For instance, with the
existing TOSA avgpool2d specification, we can not represent
`count_include_pad` and `divisor_override,` so during TorchToTosa
lowering, we silently ignore these properties which leads to SWA in some
cases—the fix captured in this change errors for unsupported scenarios.
For details on `count_include_pad` and `divisor_override,` please see
the below link.
https://pytorch.org/docs/stable/generated/torch.nn.AvgPool2d.html
---------
Co-authored-by: Hanumanth Hanumantharayappa <hhanuman@ah-hhanuman-l.dhcp.mathworks.com>
- Add/Extend Torch to TOSA legalization for the following ops:
+ Add aten.threshold_backward
+ Fix aten.threshold
+ Re-implement aten.broadcast_to using tosa.reshape and tosa.tile
+ Add support for rank 0 index for aten.index_select
+ Fix aten.index_put.hacked_twin
+ Add aten.uniform
+ Add aten.logical_and
- Update xfail_sets.py with new e2e results
- Add LIT tests to basic.mlir for newly added ops
Change-Id: I8910564a049d18293284fe2e55e82bc1d2cf10e3
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
1. Negative indices for tensor indexing is handled by wrapping around
the index values by checking their values at run time. Without the fix,
there was a runtime error.
2. Added a lit test to lock down the behavior.
3. Updated the `xfails_set` for `fx_importer_tosa` config to lockdown
the behavior with e2e test as well.
"THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY."
I've upstreamed the necessary quantized linalg Op with the
"channel-first" ordering used by torch
(https://github.com/llvm/llvm-project/pull/107740) for 2d convolution.
This patch changes the lowering for the quantized 2d case of
`aten.convolution` accordingly, which saves three transpositions per
convolution (input, weights, result) and therefore removes the
requirement to try to optimize these away in downstream passes.
Adds onnx ConvTranspose support for autopadding
(https://github.com/nod-ai/SHARK-ModelDev/issues/839).
- Adds support for attribute auto_pad="SAME_UPPER" or "SAME_LOWER" which
will automatically calculate padding of input based on output shape.
- Adds support, during auto-padding, for output_shape=[H,W] which
overrides the default output shape of input_shape[i]*stride[i] (for
spatial dimensions only).
- Adds lit test for auto-padding.
- Tests are added by https://github.com/nod-ai/SHARK-TestSuite/pull/370
NOTE: ConvTranspose still doesn't support asymmetric padding, therefore
multiple original onnx tests still won't pass.
- Add Torch to TOSA legalization for the following ops:
+ aten.empty.memory_format
+ aten.scatter.src
+ aten.slice_scatter
+ aten.diag_embed
- Update xfail_sets.py with new e2e results
- Update basic.mlir with new LIT tests
Change-Id: I817ecf207bcfcf97ca54f30c10c76c4f0f4145ae
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
Torch-to-linalg pass fails for `EmbeddingBag` when the training only
specific properties of the operator are set to `true.` For instance,
this operator's `sparse` input/property is training-specific, and if the
value of this property is `true,` the existing lowering bails out.
However, we don't need to check for training-specific parameters and
bailout from the legalization since we don't care about these properties
during the eval/inference mode.
---------
Co-authored-by: Hanumanth Hanumantharayappa <hhanuman@ah-hhanuman-l.dhcp.mathworks.com>
- Support Bidirectional LSTM (utilising the forward LSTM layer with
flipped Inputs and Outputs)
- Support layout 1
- Support default cases for attr `clip` and `input_forget`
- Support returning partial outputs (1-3)
- fixes for alt_e2e_tests lstm tests (1,2,3)
- Add Torch to TOSA lowering for aten.fill.Scalar/Tensor, aten.flip, and
aten.round
- Fix torchScalarToTosaTensor function to correctly convert Torch scalar
input to TOSA tensor
- Update xfail_sets.py with new e2e results
- Update basic.mlir with LIT tests for new ops
Change-Id: If1e42c2e582710dd8ad0465eed29806fbcdbde41
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
- Add Torch to TOSA legalization for aten.index_select
- Fix createOneDimTfIndices function in TosaLegalizeCommon.cpp to
correctly convert Torch indices to TF-style indices, which is used in
convertGatherNdOp
- Update e2e tests in xfail_sets.py
- Update basic.mlir with new LIT test for aten.index_select
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
Change-Id: I52519246183949353a3cf22f0a685fe3df8ec8ff
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
Addresses ~200 onnx model compile failures in
<https://github.com/nod-ai/SHARK-TestSuite> related to
<https://github.com/iree-org/iree/issues/18631>.
This change simplifies the result of the generated broadcast op
substantially, but reduces the case coverage slightly.
The case which will become unsupported:
- trying to actually broadcast a dynamic dim that is secretly 1.
When does this case appear in practical scenarios?
- for a model where onnx shape inference cannot figure out that a dim
should be 1.
Why do I think we should not support this case for now?
1. For all models with dynamic dim expand ops, the previous path
uniformly generates uglier linalg IR (making it harder for IREE to fuse
properly with other ops).
2. For models failing shape inference castastrophically enough to fail
to see a dim is statically 1, we can try to apply constant folding in
the onnx model before importing.
Leaving this as a draft PR, since it may be more appropriate to fix the
compilation failure in IREE rather than torch-mlir.
### Example of broadcast required in previous path:
```mlir
%300 = linalg.generic {indexing_maps = [#map11], iterator_types = ["parallel", "parallel", "parallel", "parallel"]} outs(%299 : tensor<?x12x?x?xi1>) {
^bb0(%out: i1):
%306 = linalg.index 0 : index
%307 = linalg.index 3 : index
%308 = arith.index_cast %285 : i64 to index
%309 = arith.cmpi eq, %308, %c1 : index
%310 = arith.select %309, %c0, %306 : index
%311 = arith.index_cast %286 : i64 to index
%312 = arith.cmpi eq, %311, %c1 : index
%313 = arith.select %312, %c0, %307 : index
%extracted_79 = tensor.extract %reshape_78[%310, %c0, %c0, %313] : tensor<?x1x1x?xi1>
linalg.yield %extracted_79 : i1
} -> tensor<?x12x?x?xi1>
```
### Example of broadcast with simplified shape list:
```mlir
%409 = linalg.generic {indexing_maps = [#map15, #map11], iterator_types = ["parallel", "parallel", "parallel", "parallel"]} ins(%reshape_135 : tensor<?x1x1x?xi1>) outs(%408 : tensor<?x12x?x?xi1>) {
^bb0(%in: i1, %out: i1):
linalg.yield %in : i1
} -> tensor<?x12x?x?xi1>
```
- Add lowering from Torch to TOSA for aten.diagonal
- Clean up some code
- Update xfail_sets.py with the new e2e results
- Update basic_mlir with the new op mlir test
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
Change-Id: I99bed685455752d09ed96edd837c4dfbee152701
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
- When the signal tensor is real, onnx allows its shape to be
`[batch][length]` as well as `[batch][length][1]`.
- Onnx also allows to specify `frame_length` together with `window` (not
empty), given that it matches the window size.
- Adding checks on signal and result shapes.
Previously, if the value was absent, this conversion was creating a
dense resource of value 0 with shape equal to the result shape, then
later re-extracting a splat value. This only works if the shape is
statically known, and even when the shape is known, this is completely
unnecessary since the value's shape should be `[1]` and not the result
shape.
This patch simply sets the `splatvalue` to a `torch.constant.float 0.0`
when the onnx op's `value` attr is absent, and adds `nullptr` checks to
the subsequent conditionals to avoid them in the case where an `attr` is
not given.
Addresses <https://github.com/nod-ai/SHARK-Turbine/issues/831>.
- Add Torch to TOSA legalization for the following reduction ops:
+ aten.min.dim
+ aten.min
+ aten.max
+ aten.prod
+ aten.prod.dim_int
+ aten.all.dim
- Add dtype casting support for reduce sum and prod ops
- Extend aten.max.dim legalization to a template to support aten.min.dim
legalization
- Update end-to-end tests sets in xfail_sets.py
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
Change-Id: I854dd6c0c55e570c1fb7242f20c85cf64d6e7fe0
Signed-off-by: Justin Ngo <justin.ngo@arm.com>
Bump forward and refactor inline global slots to no longer track via
symlinks. This appears to make the tests past until we manage to remove
torchscript work.
The lowering pattern for `aten.T` uses transposition implemented via
`linalg.generic`. For downstream passes it is advantageous to use named
ops wherever possible, so this patch changes the lowering to use
`linalg.transpose` instead.
This patch add a test for 638ef14, which use `linalg.broadcast` instead
of `generic` for convolution bias.
Co-authored-by: Rongsheng Gao <gaorongsheng@huawei.com>
Supports the result with dynamic shape and scalar indices like
```
func.func @test_gather_scalar(%arg0: !torch.vtensor<[3,4,5],f32>, %arg1: !torch.vtensor<[], si64>) -> !torch.vtensor<[?,?],f32> attributes {torch.onnx_meta.opset_version = 13 : si64} {
%0 = torch.operator "onnx.Gather"(%arg0, %arg1) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[3,4,5],f32>, !torch.vtensor<[], si64>) -> !torch.vtensor<[?,?],f32>
return %0 : !torch.vtensor<[?,?],f32>
}
```
`Torch::AtenSqueezeOp` is referring to the result shape, so it will
failed on lowering if the result shape is dynamic.
The `axis` attribute is optionally available. Added support by computing
the pad based on the axis values.
---------
Signed-off-by: Rob Suderman <rob.suderman@gmail.com>
- This PR adds new (and equivalent) more tensorized impl of
MelWeightMatrix which lowers all the way to linalg.
- [Ref Pytorch
Impl](https://gist.github.com/PhaneeshB/4e6dfcded3007b1b686fbe28f07a67cd)
- Thanks to @rsuderman for pointing out the difficulties [earlier
impl](#3503) posed during lowering to linalg and also for providing a
better numpy impl 🙏
This commit adds the shape info for the tensors created during the
decomposition of GroupNorm op.
Signed-Off By: Vivek Khandelwal <vivekkhandelwal1424@gmail.com>
This commit extends the OnnxToTorch lowering for BatchNormalization op
for supporting the case when training=True.
Signed-Off By: Vivek Khandelwal <vivekkhandelwal1424@gmail.com>
Giacomo Serafini
Justin Ngo
Longsheng Mou
zjgarvey
Hanumanth
aldesilv
jinchen
Sayan Saha
Sriram Kumar
Felix Schneider
David Tanner
Phaneesh Barwaria
Chi_Liu
Rob Suderman
Muhammad Abubakar
Vivek Khandelwal
Ian Wood