llvm
/
torch-mlir
mirror of https://github.com/llvm/torch-mlir
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
2,473 Commits
45 Branches
1195 Tags
69 MiB
Commit Graph

64 Commits (d3fd754b93ea10e6f3a1cc46bbb471d1a1ff287a)

Author SHA1 Message Date
Rob Suderman d3fd754b93
[onnx] `onnx.MatMulInteger` lowering to `torch.mm` and `quint*` types (#2761) 
Torch does not have an equivalent matmul operation for integers. Instead
it sidechannels the information via its quantized types. For this
lowering we setup these sidechannels then invoke `torch.mm`.
 2024-01-29 09:40:21 -08:00
Vivek Khandelwal da7c6d2c16
[MLIR][TORCH] Add support for dynamic shape for Onnx.Transpose op (#2803) 
Signed-Off By: Vivek Khandelwal <vivekkhandelwal1424@gmail.com>
 2024-01-26 09:46:54 -08:00
Phaneesh Barwaria 4964977e85
[ONNX][MLIR] support constantOfShape op (#2747) 2024-01-26 09:36:39 -08:00
Rob Suderman 60bf6c25af
[onnx] Lower `onnx.QLinearMatMul` lowering to `torch` operators (#2776) 
We can plumb the linear matmul into pytorch using its quantized types
with side channel information. To handle the final int8 operation we
dequantize and requantize.
 2024-01-24 12:28:48 -08:00
Vivek Khandelwal 894805dd5e
[MLIR][TORCH] Support for `onnx.LayerNormalization` (#2789) 
Signed-Off By: Vivek Khandelwal <vivekkhandelwal1424@gmail.com>
 2024-01-24 11:08:20 -08:00
Gaurav Shukla 12f123eff8
[ONNX][MLIR] Add support for pad op in the onnx pipeline (#2738) 
This commit adds mapping from `onnx.pad` op to `torch.pad` op. Currently
it does not support `axes` parameter of `onnx.pad` op.

Signed-off-by: Gaurav Shukla <gaurav.shukla@amd.com>
 2024-01-25 00:33:37 +05:30
Phaneesh Barwaria ac8975ea12
[MLIR] [ONNX] lowering for onnx tile op and sign op (#2725) 2024-01-24 22:56:21 +05:30
Chi_Liu 77ae56337d
[ONNX][MLIR] Add support for onnx.Exp op (#2792) 
https://github.com/nod-ai/SHARK-Turbine/issues/312
 2024-01-23 13:45:00 -08:00
James Newling dc056e58e6
[MLIR][TORCH] Add onnx.cast cases used by OPT-1.25M (#2787) 2024-01-23 21:06:25 +05:30
Gaurav Shukla b7a0329676
[ONNX][MLIR] Fix padding size constraint for onnx.maxpool op (#2782) 
Signed-off-by: Gaurav Shukla <gaurav@nod-labs.com>
 2024-01-23 19:23:01 +05:30
Chi_Liu cad98e8113
[ONNX][TORCH-MLIR] Add TopK support (#2774) 
https://github.com/nod-ai/SHARK-Turbine/issues/331
 2024-01-22 12:56:39 -08:00
Ramiro Leal-Cavazos 5883ef0f21
Fix unused variable warnings (#2775) 2024-01-22 11:05:55 -08:00
Dave Liddell 2f4924015d
[onnx] Added flatten (#2760) 
[https://github.com/nod-ai/SHARK-Turbine/issues/328](url)

---------

Co-authored-by: Dave Liddell <dliddell@xilinx.com>
 2024-01-19 16:18:16 -08:00
Gaurav Shukla 3b85c70748
[ONNX][MLIR] Add support for onnx.gather op (#2726) 
This commit adds support for gather op in the onnx pipeline.
https://github.com/nod-ai/SHARK-Turbine/issues/242

Signed-off-by: Gaurav Shukla <gaurav.shukla@amd.com>
 2024-01-19 21:58:29 +05:30
Andreas Falkenberg 4de4d38b87
Initial commit of NonZero op (#2766) 2024-01-18 15:23:13 -10:00
Rob Suderman b5387c0f29
[onnx] Lowering `onnx.dequantize_linear` to `torch` (#2759) 
We can make the per-tensor version of the operation to the dequantize
operation via marking with the make quantized tensor component. This
introductions the `qint*` and `quint*` tensor type that can be lowered
to teh appropriate dequantization behavior during the torch-to-linalg
conversion.
 2024-01-18 16:47:21 -08:00
Rob Suderman bd11877f6f
[onnx] Support lowering quantize linear to `torch` (#2751) 
We can map the per_tensor case to the `torch.aten.quantize_per_linear`
operation. In this case we extract the `scale` and `zeropoint` values
and directly invoke the quantization, then return the integer
representation value.
 2024-01-18 16:33:10 -08:00
Phaneesh Barwaria eed144bfbc
[ONNX][MLIR] add Identity op support (#2754) 2024-01-16 19:06:54 +05:30
kumardeepakamd 87389f0762
[ONNXToTorch] Add conversion for Onnx range (#2752) 
Implemented ONNX.Range. The spec says the data type for start, limit,
delta are 0-D can be double, float, int16, int32, int64, All int types
mapped to !torch.int and all float types mapped to !torch.float

---------

Co-authored-by: Kumar Deepak <kumar@xilinx.com>
 2024-01-15 14:26:46 -05:00
Rob Suderman 197b3b475c
[onnx] Convert `onnx.constant` to `torch` literal tensor (#2748) 
Handles the multiple cases of `onnx` constant values and converts them
to `torch` literal tensors. This can include splats with a single
integer or floating point value, a set of explicit integer values, or
an elements array attr of values.
 2024-01-15 09:31:22 -08:00
Chi_Liu c7452af4fa
[MLIR][ONNX] Add OnnxToTorch support for Maxpool Op (#2695) 
Add Maxpool ONNX op support.
Add Utils.h/cpp files to create a constant int list for ONNX.
 2024-01-12 14:54:38 -08:00
James Newling 47ffc90db4
signed/unsigned c++ compiler warning fixes (#2742) 2024-01-11 09:46:46 -08:00
Andreas Falkenberg 5862854bc8
[ONNX][TORCH-MLIR] LayerNorm (#2716) 
Layer Normalization using the torch.aten.native_layer_norm 

https://github.com/nod-ai/SHARK-Turbine/issues/325
 2024-01-11 14:27:04 +05:30
Xida Ren (Cedar) aee1fca251
Minor typo fix: in not implemented message for the exclusive and reverse attributes for cumsum (#2740) 2024-01-10 14:24:37 -08:00
kumardeepakamd 29569713f3
support for onnx.expand operator (#2729) 
maps onnx.expand to torch aten broadcast_to, three tests added

---------

Co-authored-by: Kumar Deepak <kumar@xilinx.com>
 2024-01-10 13:05:37 -08:00
Vivek Khandelwal 208ae35583 [MLIR][ONNX] Add TorchToOnnx Support for DepthToSpace op 
Signed-Off By: Vivek Khandelwal <vivekkhandelwal1424@gmail.com>
 2024-01-10 17:50:47 +05:30
Vivek Khandelwal 4707d3bdc6 [MLIR][ONNX] Add OnnxToTorch support for Bernoulli and CastLike op 
Signed-Off By: Vivek Khandelwal <vivekkhandelwal1424@gmail.com>
 2024-01-10 16:24:06 +05:30
Vivek Khandelwal 35e8f86792 [MLIR][ONNX] Add OnnxToTorch support for Dropout and Elu op 
Signed-Off By: Vivek Khandelwal <vivekkhandelwal1424@gmail.com>
 2024-01-10 16:23:55 +05:30
Ben Vanik 4dd17f0b71
Fixing implicit double->float truncation warnings. (#2733) 
Floating-point literals should use the correct type specifier.
 2024-01-08 17:26:38 -05:00
Han-Chung Wang 6096fcb347
[OnnxToTorch] Delete unused variables. (#2728) 2024-01-04 17:30:05 -08:00
John Wu 4e5e34d215
[MLIR][ONNX] Add OnnxToTorch support for Slice Op (#2696) 2024-01-03 19:41:10 -08:00
Xida Ren (Cedar) 1778314620
add basic cumsum. this doesn't support the exclusive and reverse attrs (#2717) 
fixes #2711
 2024-01-03 09:52:59 -08:00
Xida Ren (Cedar) 9fc212ea9a
support Onnx opset 1-13 ReduceMean where axes is supplied as an attr (#2703) 
(instead of an input)

Addresses part of #2689. fixes #2702
 2023-12-28 09:31:41 -08:00
Xida Ren (Cedar) d560698e3d
Lower `onnx.split` to `torch.aten` (#2686) 2023-12-27 17:53:07 -08:00
aldesilv 2d796b7502
lower onnx max op to torch aten maximum op (#2618) 
lower onnx min op to torch aten minimum op
 2023-12-27 11:07:35 -08:00
aldesilv 336cfb64b5
OnnxToTorch support for onnx.Mul op (#2699) 2023-12-27 10:50:08 -08:00
Xida Ren (Cedar) 6847fc1fc6
Fix since-opset too high (#2701) 
Addresses two of the ops from
https://github.com/llvm/torch-mlir/issues/2689

https://github.com/llvm/torch-mlir/issues/2700
 2023-12-27 10:08:09 -08:00
aldesilv abc6b0a25a
onnx to torch pow support (#2656) 2023-12-27 09:34:48 -08:00
Vivek Khandelwal 4f252c88b4
[MLIR][ONNX] Add OnnxToTorch support for GlobalAveragePool op. (#2692) 
This commit adds the OnnxToTorch support for GlobalAveragePool op.

Signed-Off By: vivekkhandelwal1424@gmail.com
 2023-12-26 10:25:31 -08:00
saienduri ee75e8d1ae
[MLIR][ONNX] Add OnnxToTorch support for Reshape Op (#2698) 
This commit adds the OnnxToTorch support for Reshape op.
 2023-12-26 10:20:13 -08:00
Vivek Khandelwal 0849fd0a06 [MLIR][ONNX] Fix onnx.conv lowering to handle bias tensor 
Signed-Off By: Vivek Khandelwal <vivekkhandelwal1424@gmail.com>
 2023-12-22 16:36:21 +05:30
Vivek Khandelwal 9a72c6584e [MLIR][ONNX] Add OnnxToTorch support for BatchNormalization and Concat op. 
This commit adds the OnnxToTorch support for BatchNormalization and Concat op.

Signed-Off By: vivekkhandelwal1424@gmail.com
 2023-12-22 11:25:33 +05:30
John Wu 46f2cb50dc
[onnx] Lower onnx.HardSigmoid to torch (#2682) 
The expression for HardSigmoid in Onnx
(https://onnx.ai/onnx/operators/onnx__HardSigmoid.html): max(0, min(1,
alpha * x + beta))

is inherently different from HardSigmoid in Torch
(https://pytorch.org/docs/stable/generated/torch.nn.Hardsigmoid.html)
which is: if x < -3 -> 0
elif x > 3 -> 1
else x/6 + 1/2

That being said, it was just better to compute out the entire expression
when translating the Onnx expression to Torch mlir, which is done in
this PR. Some of the logic is shared from the files in
`DecomposeComplexOps`. Therefore, refactored some shared logic between
`DecomposeComplexOps` and `DefaultDomainGToP` and put it in a `Utils`
file.
 2023-12-21 07:29:22 -08:00
Vivek Khandelwal 3226241521 [MLIR][ONNX] Add OnnxToTorch support for Conv and ConvTranspose op. 
This commit adds the OnnxToTorch support for Conv and ConvTranspose op.

Signed-Off By: vivekkhandelwal1424@gmail.com
 2023-12-21 11:12:14 +05:30
Stella Laurenzo d75cff6cd1 NFC: Remove unused variable causing a warning. 2023-12-20 19:23:27 -08:00
Rob Suderman 11cc92d4ab
[onnx] Lowerings from `onnx.tan` (#2642) 
Started work on the `tan` lowerings for ONNX to Torch. Uses `sin` and
`cos` to represent a `tan`.
 2023-12-20 10:09:39 -08:00
Andreas Falkenberg ebaab4200f
[ONNX] ONNX -> TORCH for Erf (#2673) 
TorchOnnxToTorch
For Erf function
 2023-12-19 08:07:27 -08:00
Vivek Khandelwal 8649b84e3f
[MLIR][ONNX] Add OnnxToTorch support for AveragePool op. (#2672) 
This commit adds the OnnxToTorch support for AveragePool op.

Signed-Off By: vivekkhandelwal1424@gmail.com
 2023-12-18 18:17:11 -06:00
saienduri 698ff3a736
[MLIR][ONNX] Add OnnxToTorch support for Reduction Ops (#2657) 
This commit adds the OnnxToTorch support for ReduceSum, ReduceMean, and
ReduceMin ops.
 2023-12-18 12:37:31 -08:00
John Wu deacb8ef38
[MLIR][ONNX] Add OnnxToTorch support for Gelu (#2647) 
This commit adds the OnnxToTorch support for Gelu op.

---------

Co-authored-by: Rob Suderman <suderman@google.com>
 2023-12-18 10:57:08 -08:00