Commit Graph

  • 02327af998
    Adds onnx ConvTranspose support for autopadding. (#3797) David Tanner 2024-10-18 13:31:33 -0400
  • 19ba4819ea Address CI comments Vivek Khandelwal 2024-10-18 16:44:24 +0000
  • 0fd930dff0 Disable building STABLEHLO and specify USE_MATH_DEFINES for windows builds. #3805 Sayan Saha 2024-10-18 12:12:02 -0400
  • b78fca89fc Remove passing Einsum ops from `fx_importer_stablehlo` xfail set Stephen Baione 2024-10-18 10:10:36 -0500
  • a6026b6bd0
    Merge branch 'main' into stephen-aten-_trilinear-op Stephen Baione 2024-10-18 09:49:13 -0500
  • 99d74e6bc2 Fix function description Stephen Baione 2024-10-18 09:09:49 -0500
  • 967a8b0397 [MLIR][TORCH] Add torch-onnx-to-torch-backend pipeline Vivek Khandelwal 2024-10-18 13:14:13 +0000
  • 7becc57484 Add WIP code to decompose STFT into fft or rfft giacs-epic 2024-10-18 12:19:58 +0000
  • de322ace3b Removed the redeclaration of spatial variable in MaxPool sriram 2024-10-18 14:07:33 +0530
  • 3e61ef827d shifted the padding size check above. Added test cases for maxpool-same_lower, same_upper, precomputed_same_upper sriram 2024-10-18 14:00:39 +0530
  • 9c7067649b
    build: manually update PyTorch version (#3727) Vivek Khandelwal 2024-10-18 13:32:14 +0530
  • e5f20fdf38 Emit aten.stft.center giacs-epic 2024-10-18 07:58:11 +0000
  • c58293c100 Add to appropriate xfail sets for fx_import and fx_importer_stablehlo Stephen Baione 2024-10-17 18:45:18 -0500
  • 0bc40d53c2 Add more comments zjgarvey 2024-10-17 18:43:04 -0500
  • d4a18e8e63 Merge branch 'main' of https://github.com/stbaione/torch-mlir into stephen-aten-_trilinear-op Stephen Baione 2024-10-17 17:42:14 -0500
  • 27aeb0ecbe Add trilinear to appropriate xfail sets Stephen Baione 2024-10-17 17:32:56 -0500
  • a41b939d63 Add trilinear tests to onnx xfail Stephen Baione 2024-10-17 17:17:33 -0500
  • 5fe7a9540f Simplify implementation of _trilinear, Add more test cases, Add PyTorch _trilinear "bug" to xfail set Stephen Baione 2024-10-17 17:06:14 -0500
  • 9fbed70113 update lit tests and resolve an issue with bool not being a valid tensor element type zjgarvey 2024-10-17 13:22:28 -0500
  • 0e6d87ea65 Address PR comments #3727 Vivek Khandelwal 2024-10-17 05:16:04 +0000
  • d44a848715 dont stage AtenViewOp.getSelf() zjgarvey 2024-10-16 17:32:32 -0500
  • fd7b0910ea Use SetVector instead of SmallVector zjgarvey 2024-10-16 17:10:25 -0500
  • 237326ba87 add some comments zjgarvey 2024-10-16 16:32:20 -0500
  • ea1e95fc73 Remove a problematic fold pattern for squeeze dim and add logic for handling rank 0 item ops zjgarvey 2024-10-16 16:09:49 -0500
  • 0b032f68ca Clean for readability. zjgarvey 2024-10-16 15:34:43 -0500
  • d8eadeaf50 ordered the set of ops for the pattern rewriting zjgarvey 2024-10-16 14:27:04 -0500
  • 5dc671d3e0 WIP, reimplement trilinear einsum to use AtenEinsumOp directly. Stephen Baione 2024-10-16 11:07:51 -0500
  • 8c7326f3e9 Making autopad=VALID test case more robust. #3797 David Tanner 2024-10-16 14:21:10 +0000
  • b1a776c816 Adding support for autopad=VALID and adding tests for all autopad cases. David Tanner 2024-10-16 13:57:57 +0000
  • 3319f4dc09 Fix fx_importer tests Vivek Khandelwal 2024-10-16 10:29:00 +0000
  • 367d13203e [Torch] add fold logic for some ops (#3794) yyp0 2024-10-16 16:00:58 +0800
  • dc7a1ff7d9
    [Torch] add fold logic for some ops (#3794) yyp0 2024-10-16 16:00:58 +0800
  • 13591c930c update #3794 yyp0 2024-10-16 15:28:49 +0800
  • e80016846d update testcase yyp0 2024-10-16 15:20:18 +0800
  • 66eebf7263 support div.Tensor_mode for none rounding_mode yyp0 2024-10-16 15:15:31 +0800
  • 0f52a9e9e2 update yyp0 2024-10-16 14:26:32 +0800
  • 64f965a575 Update abstract_interp_lib_gen.py Vivek Khandelwal 2024-10-15 15:43:35 +0000
  • 7df068dcfa Add more tests to xfail set Vivek Khandelwal 2024-10-15 14:24:57 +0000
  • 128b7e3a0c Update xfail set for Fx_Importer tests Vivek Khandelwal 2024-10-15 11:27:07 +0000
  • bdfacf09ee Update xfail set Vivek Khandelwal 2024-10-15 10:36:50 +0000
  • 4dd4cbdd32 Disable some filechecks due to different IRs for stable and nightly version Vivek Khandelwal 2024-10-15 10:19:41 +0000
  • 8d50d50488 Update PyTorch version Vivek Khandelwal 2024-10-15 05:45:42 +0000
  • aad34f8757 Fix lint check Vivek Khandelwal 2024-10-10 10:47:42 +0000
  • a8f8afa93f Run test for specific torch version Vivek Khandelwal 2024-10-10 10:44:29 +0000
  • a7e171fb3b Fix lint checks Vivek Khandelwal 2024-10-09 10:47:02 +0000
  • b5ffb205df This commit disables the failing sparse tensor tests since they are not maintained on day-to-day basis and blocks the roll PyTorch update for now. Vivek Khandelwal 2024-10-09 10:43:30 +0000
  • b983142ced Update torch version Vivek Khandelwal 2024-10-09 10:37:41 +0000
  • 2d4d87c976 Update filechecks for failing test Vivek Khandelwal 2024-10-07 07:15:28 +0000
  • 1eea6fec36 build: manually update PyTorch version Vivek Khandelwal 2024-10-07 06:29:58 +0000
  • 6b289f29f2
    [FxImporter] Added FxImporter test method to be executed via torch.co… (#3795) penguin_wwy 2024-10-16 10:32:52 +0800
  • dd2d305eba fix mutation importer with non persistent buffer Dixin Zhou 2024-10-15 17:32:57 -0400
  • f2fea21114 Reformatting using pre-commit clang-format. David Tanner 2024-10-15 19:41:34 +0000
  • d8a45c7c6b Merge branch 'main' of https://github.com/stbaione/torch-mlir into stephen-aten-_trilinear-op Stephen Baione 2024-10-15 14:21:31 -0500
  • d5c1774780 Adds onnx ConvTranspose support for autopadding (#839). guacamoleo 2024-10-15 18:06:47 +0000
  • b9cf6d814f Use existing AtenEinsumOp for lowering _trilinear, Lowers to torch backend, but unable to lower to linalg Stephen Baione 2024-10-15 10:46:35 -0500
  • 45bb17ebfe
    [TOSA] Add legalization for empty, scatter, slice_scatter, diag_embed (#3792) Justin Ngo 2024-10-15 08:38:02 -0700
  • 895f490cf5
    Remove checking for training specific parameters in EmbeddingBag lowering (#3782) Hanumanth04 2024-10-15 09:37:26 -0400
  • ba92091b8e update yyp0 2024-10-15 21:06:30 +0800
  • fba958c5cd update yyp0 2024-10-15 20:49:07 +0800
  • ec09892fa7 add lit testcases yyp0 2024-10-15 20:37:46 +0800
  • 8f7837d113 Ran pre-commit to fix the indendations #3782 Hanumanth Hanumantharayappa 2024-10-15 06:38:17 -0400
  • 0cff3d8233 [FxImporter] Added FxImporter test method to be executed via torch.compile backend #3795 penguin-wwy 2024-10-15 17:05:54 +0800
  • d2bdbc23fc update yyp0 2024-10-15 15:44:22 +0800
  • bc0f0e9705 update yyp0 2024-10-15 15:24:21 +0800
  • 9d85750e42 added static cast to size_t and restructured the code according to the reviewer comment sriram 2024-10-15 10:59:47 +0530
  • d23aad64b3 [Torch] add fold logic for some ops yyp0 2024-10-15 11:58:05 +0800
  • d3b29dafe5 temp zjgarvey 2024-10-14 20:08:53 -0500
  • c630025658 [TOSA] Add legalization for empty, scatter, slice_scatter, diag_embed ops #3792 Justin Ngo 2024-10-14 20:45:21 +0000
  • 1e431c6a90
    Add AtenSliceTOp Canonicalization to SimplifyShapeCalculations pass (#3791) zjgarvey 2024-10-14 14:41:31 -0500
  • 768e0de8d5 Push with changes with pre-commit setup Hanumanth Hanumantharayappa 2024-10-14 15:20:30 -0400
  • e2be336c08 Move checks to be consistent with rest of file #3791 zjgarvey 2024-10-14 14:19:40 -0500
  • 09c6f835f0 Add AtenSliceTOp Canonicalization to SimplifyShapeCalculations pass zjgarvey 2024-10-14 13:06:33 -0500
  • edd1bbec46
    Integrate LLVM at llvm/llvm-project@c13f806 (#3789) Marius Brehler 2024-10-14 15:00:45 +0200
  • 26f03ce62a
    Integrate LLVM at llvm/llvm-project@c13f806 #3789 Marius Brehler 2024-10-14 10:01:48 +0000
  • 896bcd3cf7 Add NoMemoryEffect for ops with value semantics memory_effect Ian Wood 2024-10-13 15:06:32 -0500
  • 33ad5ff155 [Torch] support 1d aten tensor shape and dtype infer (#3776) yyp0 2024-10-12 17:51:15 +0800
  • fa26bfc0d6 [Torch] support adaptive_max_pool1d when return_indices equals False (#3783) yyp0 2024-10-11 23:42:15 +0800
  • b176939808
    [Torch] support 1d aten tensor shape and dtype infer (#3776) yyp0 2024-10-12 17:51:15 +0800
  • 2e44fa0cb3 Bump to llvm/llvm-project@c13f806f17 #3786 Ubuntu 2024-10-11 20:08:09 +0000
  • ab62f35373
    Add more patterns to scalarize-shapes pass (#3781) zjgarvey 2024-10-11 11:15:17 -0500
  • 7b11dfc0ee
    [Torch] support adaptive_max_pool1d when return_indices equals False (#3783) yyp0 2024-10-11 23:42:15 +0800
  • 240104ea4d Generate aten.mul.left_t (to simplify shape inference of aten.tile) giacs-epic 2024-10-11 13:59:28 +0000
  • cfe34ba88e Remove left over debug print statements Stephen Baione 2024-10-11 08:54:38 -0500
  • f2a4e9bb52 Current state Stephen Baione 2024-10-11 08:51:48 -0500
  • a4f8215df9 update #3783 yyp0 2024-10-11 19:50:50 +0800
  • cc7e8e72a2 update yyp0 2024-10-11 19:44:21 +0800
  • 33d989bae9 Fix to remove checking for training specific parameters in EmbeddedingBag lowering Hanumanth Hanumantharayappa 2024-10-11 06:54:17 -0400
  • 8787970afe
    [Torch] Fold no-op reshape (#3769) Ian Wood 2024-10-10 18:54:27 -0700
  • 7f8418f91f Add another test #3781 zjgarvey 2024-10-10 20:47:45 -0500
  • 6550542a14 add a few tests zjgarvey 2024-10-10 20:37:47 -0500
  • 357db3e638 Add patterns for propagating shapes through AtenWhereSelf and AtenEqTensor Add fold pattern for a rank0 squeezeDim of a full op Add support for getting a list from a splat ValueTensorLiteralOp for materializing scalar comparisons in where.self and eq.tensor zjgarvey 2024-10-10 17:41:26 -0500
  • e3ff689a96 Add description of method Stephen Baione 2024-10-10 17:15:49 -0500
  • 9db2ea4306 Add op to `populateLinearPatternsAndLegality` Stephen Baione 2024-10-10 16:57:58 -0500
  • ee7f6ee9fd
    Merge branch 'main' into lower_torch_aten_gcd_to_linalg_and_scf #3732 bratislavSyrmia 2024-10-10 22:45:58 +0200
  • f84cc4984f CHECKPOINT: Initial implementation of ate::_trilinear, which implements a trilinear einstein sum. WIP, it currently builds, but fails at lowering to linalg Stephen Baione 2024-10-10 15:03:28 -0500
  • 5c9f369b17 Fix translation of col2im shape function giacs-epic 2024-10-10 16:51:03 +0000
  • bda2214d49 Rewrite get_registered_ops.cpp to python version #3780 AmosLewis 2024-10-10 08:53:46 -0700
  • 2665ed343b
    adds a few common patterns to scalarize shapes pass (#3779) zjgarvey 2024-10-10 10:16:45 -0500
  • e75c870f1d Add torch.aten.eq.bool giacs-epic 2024-10-10 11:18:28 +0000
  • 52ecff831b [stablehlo] support aten.view.dtype lowering (#3778) yyp0 2024-10-10 15:50:17 +0800