Commit Graph

  • 9be3b6c578 lint zjgarvey 2024-10-30 15:06:52 -0500
  • 90789b8a29 fix bug in transpose propagation zjgarvey 2024-10-30 14:47:26 -0500
  • 873a49ab6d [TOSA] Expand Torch to TOSA legalization coverage #3827 Justin Ngo 2024-10-29 18:55:18 +0000
  • a6292f38ca
    [bazel] Fix missing dependency in the build (#3826) Rob Suderman 2024-10-30 11:47:04 -0700
  • 4f0818fb71 Add some tests, remove uneccessary patterns zjgarvey 2024-10-30 13:00:04 -0500
  • 740a1238c6 [bazel] Fix missing dependency in the build #3826 Rob Suderman 2024-10-29 16:02:55 -0700
  • 7bf2597193 Changed calls to yaml.load to explicitly use the safe loader. #3817 TimAtGoogle 2024-10-29 18:01:52 -0500
  • 7fd98efca8 Augmented calls to yaml.load to use the safe loader. TimAtGoogle 2024-10-29 17:37:17 -0500
  • 6f34294dc9 Augmented calls to yaml.load to use the safe loader. TimAtGoogle 2024-10-24 15:47:30 -0500
  • 95623e7ccc Add a propagation pattern for transpose.int zjgarvey 2024-10-30 11:41:43 -0500
  • 033f384bd5 [chore]: remove unused variable wujiawei.aml 2024-10-30 16:29:19 +0800
  • c65996c594 [stablehlo] fix: enhance torch's index-like op lowering to stablehloe's gather/scatter wujiawei.aml 2024-10-30 14:55:56 +0800
  • 8b0bf2e293
    Bump LLVM to llvm/llvm-project@6c64c8a6f3 (#3818) Max191 2024-10-30 11:38:51 -0400
  • 6b58c89914
    Remove variable used for only assertion (#3837) Max191 2024-10-30 10:51:06 -0400
  • 76e57d6cc9 Remove variable used for only assertion #3837 Max Dawkins 2024-10-30 10:07:23 -0400
  • d139d142a7 bump stablehlo #3818 Max Dawkins 2024-10-25 15:36:34 -0400
  • 517cbc66f4 Bump LLVM to llvm/llvm-project@6c64c8a6f3 Max Dawkins 2024-10-24 21:42:59 -0400
  • 16b3bd6e6c
    build: manually update PyTorch version and fix CI failure (#3830) Vivek Khandelwal 2024-10-30 18:56:01 +0530
  • d45ccc6765
    Merge branch 'main' into stephen-aten-_trilinear-op #3784 Stephen Baione 2024-10-30 08:13:01 -0500
  • 9ab2a150f2
    [Torch] emit upsample_bilinear2d(.vec) ops (#3834) Yuanqiang Liu 2024-10-30 20:18:24 +0800
  • 912a2c0231 update #3834 Yuanqiang Liu 2024-10-30 20:04:40 +0800
  • ebf577b127 Update dtype check #3830 Vivek Khandelwal 2024-10-30 11:46:20 +0000
  • 65e19f0f94 update Yuanqiang Liu 2024-10-30 19:05:38 +0800
  • c6688fbcd0 Update xfail set for fx_importer_stablehlo Vivek Khandelwal 2024-10-30 10:55:56 +0000
  • 3d11dccd63 fix tests that fail on nightly #3835 Andrija Bosnjakovic 2024-10-30 11:12:35 +0100
  • f7c78c53b8 Update xfail set Vivek Khandelwal 2024-10-30 10:38:08 +0000
  • e467a47859 update xfail yyp0 2024-10-30 17:56:08 +0800
  • dab9b54dd8 format Yuanqiang Liu 2024-10-30 17:40:38 +0800
  • 78ee7ff1d1 [Torch] emit upsample_bilinear2d(.vec) ops Yuanqiang Liu 2024-10-30 17:36:15 +0800
  • 35f184fe37 update yyp0 2024-10-30 16:46:48 +0800
  • 5d01bc7e19 [Torch] support AtenNllLossForwardOp decomposition yyp0 2024-10-30 16:43:43 +0800
  • 45ae63ce32 [Torch] support AtenExp2Op yyp0 2024-10-30 16:34:19 +0800
  • 659a18f7c9 Roll PyTorch draft Vivek Khandelwal 2024-10-30 08:17:14 +0000
  • 419b3e2b6e Changed calls to yaml.load to explicitly use the safe loader. #3825 fixing_security_issues TimAtGoogle 2024-10-29 18:01:52 -0500
  • 58041a4713 Augmented calls to yaml.load to use the safe loader. TimAtGoogle 2024-10-29 17:37:17 -0500
  • 25951c4e1e Augmented calls to yaml.load to use the safe loader. TimAtGoogle 2024-10-24 15:47:30 -0500
  • 8be3bcb551 Add scalarization for view like ops and more complex slices zjgarvey 2024-10-29 17:03:02 -0500
  • adc99e2046 Fixed a typo Hanumanth Hanumantharayappa 2024-10-29 14:12:44 -0400
  • a7b461fccf Add lit tests zjgarvey 2024-10-29 12:04:08 -0500
  • 2f4448f89e Add unit tests to lock-down the changes Hanumanth Hanumantharayappa 2024-10-29 12:03:05 -0400
  • fb459ac33a Add arith lowering for binary bool ops zjgarvey 2024-10-29 10:30:04 -0500
  • 8f3826a36d Fix torchToTosa lowering for avgpool2d to handle unsupported parameters Hanumanth Hanumantharayappa 2024-10-29 09:44:58 -0400
  • 2b01f8b7f3
    [Tosa] : Add support for negative indices in index.tensor and index.Tensor_hacked_twin for TorchToTosa lowering. (#3790) Sayan Saha 2024-10-25 18:37:19 -0400
  • 5003cefcf4 Merge branch 'main' of https://github.com/stbaione/torch-mlir into stephen-aten-_trilinear-op Stephen Baione 2024-10-25 12:16:56 -0500
  • 54d9e24013
    [TorchToLinalg] Implement lowering of torch.aten.rrelu_with_noise and torch.aten.rrelu_with_noise_backward ops (fix) (#3748) Andrija Bosnjakovic 2024-10-25 18:01:05 +0200
  • ad9dfe974e
    Fix clang warning about printf format (#3814) Dmitry Babokin 2024-10-25 00:42:08 -0700
  • 76209db5a5
    Update quantized matmul tests to DQ/Q format supported by fx_importer (#3815) Felix Schneider 2024-10-24 21:59:58 +0200
  • 7cf1bb9a93 Add test with pytorch fix coming in #3733 jinchen62 2024-10-24 12:45:17 -0700
  • 443b5cd2e1 Generalize MaxUnpool lowering including 2d case jinchen62 2024-10-08 18:04:32 -0700
  • 76e636dc71 Generalize max_unpool lowering jinchen62 2024-09-25 07:56:10 -0700
  • 1259e8a00a
    Add Some Folders For Small Reshape Ops (#3813) zjgarvey 2024-10-24 12:09:00 -0500
  • ca9987a201 fix slicedStrides problem found by lionsky #3721 Cedar 2024-10-18 10:08:54 -0700
  • 0055336a56 add test cases and fix err Xida Ren 2024-10-03 16:29:28 +0000
  • 1e376d23c7 use ScatterReduceTwo instead because ScatterReduce fails for unknown reasons Xida Ren 2024-09-25 16:14:01 +0000
  • 9f8e3ebf11 return indices in correct multidimensional format Xida Ren 2024-09-24 16:50:16 +0000
  • 868a6bd851 initial indexing fix Xida Ren 2024-09-23 19:51:05 +0000
  • 62894bcc7d rewrite the decomposition and fix a million type issues Xida Ren 2024-09-20 18:19:40 +0000
  • 3c2c4c1852 modify comment formatting #3813 zjgarvey 2024-10-24 11:38:10 -0500
  • ae145aa73e Add simplifications to unroll+fold list generation giacs-epic 2024-10-24 16:36:36 +0000
  • 8b70b602ab Update quantized matmul tests to DQ/Q format supported by fx_importer #3815 Felix Schneider 2024-10-24 13:07:45 +0200
  • 5c080f1c91 Fix clang warning about printf format #3814 Dmitry Babokin 2024-10-23 18:23:06 -0700
  • efea881dc5 Rebase main + update xfail_sets #3790 Sayan Saha 2024-10-16 19:05:44 -0400
  • 9e4962c284 [TOSA] : Add support for negative indices for torch.index.tensor Sayan Saha 2024-10-11 12:08:34 -0400
  • d6feb2179c
    Added support for Maxpool (Autopad) (#3774) Sriram Kumar 2024-10-23 18:34:50 +0530
  • 2f9a68cc1e
    Add canonicalization pattern for maxpool3d with indices op (#3704) lingzhiz1998 2024-10-23 21:01:20 +0800
  • 55ff110dc2
    [MLIR][TORCH] Only unroll prim loop-like ops within a `torch.shape.calculate` region (#3812) zjgarvey 2024-10-23 03:08:55 -0500
  • fab08fdef7 black zjgarvey 2024-10-22 21:25:54 -0500
  • 02da96f29e update xfails zjgarvey 2024-10-22 21:22:15 -0500
  • cdd64f5f66 Add lit tests, expand transpose noop fold zjgarvey 2024-10-22 21:12:11 -0500
  • 381fe26330 add much needed folders for pad mess coming from torch->ONNX->torch path zjgarvey 2024-10-22 19:43:07 -0500
  • 3f5847d789 update comments, add a nullptr check, and add a test #3812 zjgarvey 2024-10-22 14:58:51 -0500
  • aca33f1742
    [TorchToLinalg] Use Op with native channel order for quantized conv2d (#3807) Felix Schneider 2024-10-22 20:26:16 +0200
  • 065b36c59f Only unroll prim loops ops within a shape.calculate region zjgarvey 2024-10-22 13:25:21 -0500
  • 688b17824a [TorchToLinalg] Use Op with native channel order for quantized conv2d #3807 Felix Schneider 2024-10-20 19:37:38 +0200
  • 42ba541c68
    [fx] Fix importing and tests for quantized conv (#3809) Felix Schneider 2024-10-22 18:37:57 +0200
  • 29897eacfc remove the now superfluous bias scaling #3809 Felix Schneider 2024-10-22 18:04:28 +0200
  • 0f5a2dc844 [uncategorized_lowerings] Add lowering for torch.aten.round.decimals #3811 meshtag 2024-10-22 10:54:26 +0530
  • 140cad5659
    Add More Scalarize Shapes Patterns (#3810) zjgarvey 2024-10-21 19:42:39 -0500
  • 95603f60c1 fix lit test #3810 zjgarvey 2024-10-21 19:12:36 -0500
  • 38c23edcd2 add two more anchor ops, fold squeeze dim, improve scalar arithmetic folders zjgarvey 2024-10-21 19:03:37 -0500
  • 32f4140a0d Add a size check and some arithmetic scalarization patterns zjgarvey 2024-10-21 17:08:44 -0500
  • 6e9e4b9443 Add a folding pattern for AtenEqIntOp zjgarvey 2024-10-21 14:51:26 -0500
  • b85d8a5bed Add a pattern for broadcast zjgarvey 2024-10-21 14:19:17 -0500
  • f5048dcffe Make bias an integer Felix Schneider 2024-10-21 21:04:45 +0200
  • a83e106f92
    Rework Scalarize Shapes Pass (#3799) zjgarvey 2024-10-21 12:47:19 -0500
  • dc8dec4131 rework control flow in getListFromTensor for readability #3799 zjgarvey 2024-10-21 12:36:13 -0500
  • df3e3ab45d Add Q op after f32 operator Felix Schneider 2024-10-21 19:02:55 +0200
  • 4f4fa4a0cb Merge remote-tracking branch 'upstream/main' into scalarize_shapes_rework zjgarvey 2024-10-21 11:22:44 -0500
  • fa4794dae2
    [MLIR][TORCH] Add torch-onnx-to-torch-backend pipeline (#3801) Vivek Khandelwal 2024-10-21 21:50:44 +0530
  • 34af8330e8 Further decompose stack into unsqueeze+cat giacs-epic 2024-10-21 15:57:44 +0000
  • 98ca822e5d [fx] Fix importing and tests for quantized conv Felix Schneider 2024-10-21 16:20:22 +0200
  • d2330df58f
    build: manually update PyTorch version (#3808) Vivek Khandelwal 2024-10-21 17:26:09 +0530
  • 63d9818f6e build: manually update PyTorch version #3808 Vivek Khandelwal 2024-10-21 11:43:42 +0000
  • ccc65fe2c9 Address CI comment #3801 Vivek Khandelwal 2024-10-21 08:11:25 +0000
  • f5d15ab20e
    Bump LLVM to llvm/llvm-project@f0b3b6d1 (#3806) Felix Schneider 2024-10-20 18:32:21 +0200
  • 2aeb4f1485 Bump https://github.com/llvm/llvm-project/commit/f0b3b6d1 #3806 Felix Schneider 2024-10-20 16:23:20 +0200
  • 90df5464ef Removed the padding tensor satisfy constraint code #3774 sriram 2024-10-19 09:14:01 +0530
  • 47312f8936 Rework for better folding and accessing value tensor literal data 1. getListFromTensor now modifies a vector of `OpFoldResult` 2. added a helper function for writing an `aten.Tensor` op to clean up code 3. other misc cleaning of propagation patterns. 4. adds a lit test which demonstates the changes to getListFromTensor rework 5. add a few more comments zjgarvey 2024-10-18 15:42:41 -0500
  • 4df4f16e0e Add check for case with a triple intersecting expand dimension that is not included in sumDim, Add note in func description that `unrollDim` is unused Stephen Baione 2024-10-18 14:17:39 -0500
  • 09cdbe4c47
    Disable building STABLEHLO and specify USE_MATH_DEFINES for windows builds. (#3805) Sayan Saha 2024-10-18 15:04:37 -0400