llvm
/
torch-mlir
mirror of https://github.com/llvm/torch-mlir
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
torch-mlir/test/Conversion/TorchToStablehlo/pooling.mlir

220 lines
18 KiB
MLIR
Raw Normal View History

mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// RUN: torch-mlir-opt <%s -convert-torch-to-stablehlo -split-input-file -verify-diagnostics | FileCheck %s
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
// -----
// CHECK-LABEL: func.func @torch.aten.max_pool2d(
// CHECK-SAME: %[[VAL_0:.*]]: !torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[?,?,?,?],f32> {
// CHECK: %[[VAL_1:.*]] = torch_c.to_builtin_tensor %[[VAL_0]] : !torch.vtensor<[?,?,?,?],f32> -> tensor<?x?x?x?xf32>
// CHECK: %int2 = torch.constant.int 2
// CHECK: %int1 = torch.constant.int 1
// CHECK: %int0 = torch.constant.int 0
// CHECK: %false = torch.constant.bool false
// CHECK: %[[VAL_2:.*]] = torch.prim.ListConstruct %int2, %int2 : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[VAL_3:.*]] = torch.prim.ListConstruct %int1, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[VAL_4:.*]] = torch.prim.ListConstruct %int0, %int0 : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[VAL_5:.*]] = torch.prim.ListConstruct %int2, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
[Stablehlo] fix reduce max init_value with -inf (#2064) * [Stablehlo] fix reduce max init_value with -inf * update
2023-05-07 03:05:51 +08:00
// CHECK: %[[VAL_6:.*]] = stablehlo.constant dense<0xFF800000> : tensor<f32>
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[VAL_7:.*]] = "stablehlo.reduce_window"(%[[VAL_1]], %[[VAL_6]]) ({
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
// CHECK: ^bb0(%[[VAL_8:.*]]: tensor<f32>, %[[VAL_9:.*]]: tensor<f32>):
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[VAL_10:.*]] = stablehlo.maximum %[[VAL_8]], %[[VAL_9]] : tensor<f32>
// CHECK: stablehlo.return %[[VAL_10]] : tensor<f32>
Bump stablehlo to openxla/stablehlo@fd52182f76cadb82f2064fe5fc49a4fb4347a826 (#2821) With the recent LLVM integrate and changes from https://github.com/llvm/llvm-project/pull/78260, we hit this build error in Stablehlo (which is quite old). ``` external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1020:14: error: no member named 'startRootUpdate' in 'mlir::PatternRewriter' rewriter.startRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1026:16: error: no member named 'finalizeRootUpdate' in 'mlir::PatternRewriter' rewriter.finalizeRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1029:16: error: no member named 'cancelRootUpdate' in 'mlir::PatternRewriter' rewriter.cancelRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1108:14: error: no member named 'updateRootInPlace' in 'mlir::PatternRewriter' rewriter.updateRootInPlace(op->getParentOp(), [&]() { return; }); ~~~~~~~~ ^ 4 errors generated. Target @torch-mlir//:torch-mlir-opt failed to build ``` I'm still puzzled as to how this didn't fail with the CMake merge gating CI (do we not test Stablehlo builds/tests?). In any case, bumping our submodule to https://github.com/openxla/stablehlo/pull/1918 fixes it. It exposes a new failing lit test in TorchToStablehlo though, that I have looped stablehlo developers into ([here](https://discord.com/channels/999073994483433573/999074539138990131/1201235845391331419)). ``` bazel run @torch-mlir//test/Conversion:TorchToStablehlo/scatter.mlir.test ...external/torch-mlir/test/Conversion/TorchToStablehlo/scatter.mlir within split at <stdin>:1 offset :33:8: error: unexpected error: Expects non-empty reduction block for type inference %0 = torch.aten.scatter.src %arg0, %int0, %arg1, %arg2 : !torch.vtensor<[?,?],si64>, !torch.int, !torch.vtensor<[?,?],si64>, !torch.vtensor<[?,?],si64> -> !torch.vtensor<[?,?],si64> ^ LLVM ERROR: Failed to infer result type(s). ``` Bazel CI: https://github.com/sjain-stanford/torch-mlir/actions/runs/7732673480/job/21083102228
2024-02-01 06:21:17 +08:00
// CHECK: }) {padding = dense<0> : tensor<4x2xi64>, window_dilations = array<i64: 1, 1, 2, 1>, window_dimensions = array<i64: 1, 1, 2, 2>, window_strides = array<i64: 1, 1, 1, 1>} : (tensor<?x?x?x?xf32>, tensor<f32>) -> tensor<?x?x?x?xf32>
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
// CHECK: %[[VAL_11:.*]] = torch_c.from_builtin_tensor %[[VAL_7]] : tensor<?x?x?x?xf32> -> !torch.vtensor<[?,?,?,?],f32>
// CHECK: return %[[VAL_11]] : !torch.vtensor<[?,?,?,?],f32>
func.func @torch.aten.max_pool2d(%arg0: !torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[?,?,?,?],f32> {
%int2 = torch.constant.int 2
%int1 = torch.constant.int 1
%int0 = torch.constant.int 0
%false = torch.constant.bool false
%0 = torch.prim.ListConstruct %int2, %int2 : (!torch.int, !torch.int) -> !torch.list<int>
%1 = torch.prim.ListConstruct %int1, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
%2 = torch.prim.ListConstruct %int0, %int0 : (!torch.int, !torch.int) -> !torch.list<int>
%3 = torch.prim.ListConstruct %int2, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
%4 = torch.aten.max_pool2d %arg0, %0, %1, %2, %3, %false : !torch.vtensor<[?,?,?,?],f32>, !torch.list<int>, !torch.list<int>, !torch.list<int>, !torch.list<int>, !torch.bool -> !torch.vtensor<[?,?,?,?],f32>
return %4 : !torch.vtensor<[?,?,?,?],f32>
}
// -----
// CHECK-LABEL: func.func @torch.aten.max_pool2d$padding(
// CHECK-SAME: %[[VAL_0:.*]]: !torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[?,?,?,?],f32> {
// CHECK: %[[VAL_1:.*]] = torch_c.to_builtin_tensor %[[VAL_0]] : !torch.vtensor<[?,?,?,?],f32> -> tensor<?x?x?x?xf32>
// CHECK: %int2 = torch.constant.int 2
// CHECK: %int1 = torch.constant.int 1
// CHECK: %false = torch.constant.bool false
// CHECK: %[[VAL_2:.*]] = torch.prim.ListConstruct %int2, %int2 : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[VAL_3:.*]] = torch.prim.ListConstruct %int1, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[VAL_4:.*]] = torch.prim.ListConstruct %int2, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
[Stablehlo] fix reduce max init_value with -inf (#2064) * [Stablehlo] fix reduce max init_value with -inf * update
2023-05-07 03:05:51 +08:00
// CHECK: %[[VAL_5:.*]] = stablehlo.constant dense<0xFF800000> : tensor<f32>
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[VAL_6:.*]] = "stablehlo.reduce_window"(%[[VAL_1]], %[[VAL_5]]) ({
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
// CHECK: ^bb0(%[[VAL_8:.*]]: tensor<f32>, %[[VAL_9:.*]]: tensor<f32>):
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[VAL_10:.*]] = stablehlo.maximum %[[VAL_8]], %[[VAL_9]] : tensor<f32>
// CHECK: stablehlo.return %[[VAL_10]] : tensor<f32>
[torch-mlir][conversion-test] cleanup trailing whitespace in mlir files (#2807)
2024-01-26 06:24:28 +08:00
// CHECK: })
Bump stablehlo to openxla/stablehlo@fd52182f76cadb82f2064fe5fc49a4fb4347a826 (#2821) With the recent LLVM integrate and changes from https://github.com/llvm/llvm-project/pull/78260, we hit this build error in Stablehlo (which is quite old). ``` external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1020:14: error: no member named 'startRootUpdate' in 'mlir::PatternRewriter' rewriter.startRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1026:16: error: no member named 'finalizeRootUpdate' in 'mlir::PatternRewriter' rewriter.finalizeRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1029:16: error: no member named 'cancelRootUpdate' in 'mlir::PatternRewriter' rewriter.cancelRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1108:14: error: no member named 'updateRootInPlace' in 'mlir::PatternRewriter' rewriter.updateRootInPlace(op->getParentOp(), [&]() { return; }); ~~~~~~~~ ^ 4 errors generated. Target @torch-mlir//:torch-mlir-opt failed to build ``` I'm still puzzled as to how this didn't fail with the CMake merge gating CI (do we not test Stablehlo builds/tests?). In any case, bumping our submodule to https://github.com/openxla/stablehlo/pull/1918 fixes it. It exposes a new failing lit test in TorchToStablehlo though, that I have looped stablehlo developers into ([here](https://discord.com/channels/999073994483433573/999074539138990131/1201235845391331419)). ``` bazel run @torch-mlir//test/Conversion:TorchToStablehlo/scatter.mlir.test ...external/torch-mlir/test/Conversion/TorchToStablehlo/scatter.mlir within split at <stdin>:1 offset :33:8: error: unexpected error: Expects non-empty reduction block for type inference %0 = torch.aten.scatter.src %arg0, %int0, %arg1, %arg2 : !torch.vtensor<[?,?],si64>, !torch.int, !torch.vtensor<[?,?],si64>, !torch.vtensor<[?,?],si64> -> !torch.vtensor<[?,?],si64> ^ LLVM ERROR: Failed to infer result type(s). ``` Bazel CI: https://github.com/sjain-stanford/torch-mlir/actions/runs/7732673480/job/21083102228
2024-02-01 06:21:17 +08:00
// CHECK-SAME{LITERAL}: {padding = dense<[[0, 0], [0, 0], [2, 2], [1, 1]]> : tensor<4x2xi64>, window_dilations = array<i64: 1, 1, 2, 1>, window_dimensions = array<i64: 1, 1, 2, 2>, window_strides = array<i64: 1, 1, 1, 1>} : (tensor<?x?x?x?xf32>, tensor<f32>) -> tensor<?x?x?x?xf32>
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
// CHECK: %[[VAL_7:.*]] = torch_c.from_builtin_tensor %[[VAL_6]] : tensor<?x?x?x?xf32> -> !torch.vtensor<[?,?,?,?],f32>
// CHECK: return %[[VAL_7]] : !torch.vtensor<[?,?,?,?],f32>
func.func @torch.aten.max_pool2d$padding(%arg0: !torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[?,?,?,?],f32> {
%int2 = torch.constant.int 2
%int1 = torch.constant.int 1
%false = torch.constant.bool false
%0 = torch.prim.ListConstruct %int2, %int2 : (!torch.int, !torch.int) -> !torch.list<int>
%1 = torch.prim.ListConstruct %int1, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
%2 = torch.prim.ListConstruct %int2, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
%3 = torch.aten.max_pool2d %arg0, %0, %1, %2, %2, %false : !torch.vtensor<[?,?,?,?],f32>, !torch.list<int>, !torch.list<int>, !torch.list<int>, !torch.list<int>, !torch.bool -> !torch.vtensor<[?,?,?,?],f32>
return %3 : !torch.vtensor<[?,?,?,?],f32>
}
// -----
build: update llvm tag to 74fb770d (#1539) * build: update llvm tag to 74fb770d This commit makes the following changes needed to update bump LLVM: + replace usages of `tensor::createPadScalarOp`, see https://reviews.llvm.org/D136493 + Update file checks
2022-11-01 15:27:09 +08:00
// CHECK-LABEL: func.func @torch.aten.max_pool2d_with_indices(
// CHECK-SAME: %[[ARG0:.*]]: !torch.vtensor<[?,?,?],f32>) -> (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[?,?,?],si64>) {
// CHECK: %[[T0:.*]] = torch_c.to_builtin_tensor %[[ARG0]] : !torch.vtensor<[?,?,?],f32> -> tensor<?x?x?xf32>
// CHECK: %[[INT3:.*]] = torch.constant.int 3
// CHECK: %[[INT2:.*]] = torch.constant.int 2
// CHECK: %[[FALSE:.*]] = torch.constant.bool false
// CHECK: %[[INT0:.*]] = torch.constant.int 0
// CHECK: %[[INT1:.*]] = torch.constant.int 1
// CHECK: %[[T1:.*]] = torch.prim.ListConstruct %[[INT3]], %[[INT3]] : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[T2:.*]] = torch.prim.ListConstruct %[[INT2]], %[[INT2]] : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[T3:.*]] = torch.prim.ListConstruct %[[INT0]], %[[INT0]] : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[T4:.*]] = torch.prim.ListConstruct %[[INT1]], %[[INT1]] : (!torch.int, !torch.int) -> !torch.list<int>
[Stablehlo] fix reduce max init_value with -inf (#2064) * [Stablehlo] fix reduce max init_value with -inf * update
2023-05-07 03:05:51 +08:00
// CHECK: %[[T5:.*]] = stablehlo.constant dense<0xFF800000> : tensor<f32>
build: update llvm tag to 74fb770d (#1539) * build: update llvm tag to 74fb770d This commit makes the following changes needed to update bump LLVM: + replace usages of `tensor::createPadScalarOp`, see https://reviews.llvm.org/D136493 + Update file checks
2022-11-01 15:27:09 +08:00
// CHECK: %[[C0:.*]] = arith.constant 0 : index
// CHECK: %[[DIM:.*]] = tensor.dim %[[T0]], %[[C0]] : tensor<?x?x?xf32>
// CHECK: %[[T6:.*]] = arith.index_cast %[[DIM]] : index to i64
// CHECK: %[[C1:.*]] = arith.constant 1 : index
// CHECK: %[[DIM_0:.*]] = tensor.dim %[[T0]], %[[C1]] : tensor<?x?x?xf32>
// CHECK: %[[T7:.*]] = arith.index_cast %[[DIM_0]] : index to i64
// CHECK: %[[C2:.*]] = arith.constant 2 : index
// CHECK: %[[DIM_1:.*]] = tensor.dim %[[T0]], %[[C2]] : tensor<?x?x?xf32>
// CHECK: %[[T8:.*]] = arith.index_cast %[[DIM_1]] : index to i64
// CHECK: %[[FROM_ELEMENTS:.*]] = tensor.from_elements %[[T6]], %[[T7]], %[[T8]] : tensor<3xi64>
// CHECK: %[[T9:.*]] = arith.muli %[[T8]], %[[T7]] : i64
// CHECK: %[[FROM_ELEMENTS_2:.*]] = tensor.from_elements %[[T6]], %[[T9]] : tensor<2xi64>
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[T10:.*]] = stablehlo.dynamic_iota %[[FROM_ELEMENTS_2]], dim = 1 : (tensor<2xi64>) -> tensor<?x?xi64>
// CHECK: %[[T11:.*]] = stablehlo.dynamic_reshape %[[T10]], %[[FROM_ELEMENTS]] : (tensor<?x?xi64>, tensor<3xi64>) -> tensor<?x?x?xi64>
// CHECK: %[[T12:.*]] = stablehlo.constant dense<0> : tensor<i64>
// CHECK: %[[T13:.*]]:2 = "stablehlo.reduce_window"(%[[T0]], %[[T11]], %[[T5]], %[[T12]]) ({
build: update llvm tag to 74fb770d (#1539) * build: update llvm tag to 74fb770d This commit makes the following changes needed to update bump LLVM: + replace usages of `tensor::createPadScalarOp`, see https://reviews.llvm.org/D136493 + Update file checks
2022-11-01 15:27:09 +08:00
// CHECK: ^bb0(%[[ARG1:.*]]: tensor<f32>, %[[ARG2:.*]]: tensor<i64>, %[[ARG3:.*]]: tensor<f32>, %[[ARG4:.*]]: tensor<i64>):
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[T16:.*]] = stablehlo.compare GE, %[[ARG1]], %[[ARG3]], FLOAT : (tensor<f32>, tensor<f32>) -> tensor<i1>
// CHECK: %[[T17:.*]] = stablehlo.select %[[T16]], %[[ARG1]], %[[ARG3]] : tensor<i1>, tensor<f32>
// CHECK: %[[T18:.*]] = stablehlo.compare EQ, %[[ARG1]], %[[ARG3]], FLOAT : (tensor<f32>, tensor<f32>) -> tensor<i1>
// CHECK: %[[T19:.*]] = stablehlo.minimum %[[ARG2]], %[[ARG4]] : tensor<i64>
// CHECK: %[[T20:.*]] = stablehlo.select %[[T16]], %[[ARG2]], %[[ARG4]] : tensor<i1>, tensor<i64>
// CHECK: %[[T21:.*]] = stablehlo.select %[[T18]], %[[T19]], %[[T20]] : tensor<i1>, tensor<i64>
// CHECK: stablehlo.return %[[T17]], %[[T21]] : tensor<f32>, tensor<i64>
Bump stablehlo to openxla/stablehlo@fd52182f76cadb82f2064fe5fc49a4fb4347a826 (#2821) With the recent LLVM integrate and changes from https://github.com/llvm/llvm-project/pull/78260, we hit this build error in Stablehlo (which is quite old). ``` external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1020:14: error: no member named 'startRootUpdate' in 'mlir::PatternRewriter' rewriter.startRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1026:16: error: no member named 'finalizeRootUpdate' in 'mlir::PatternRewriter' rewriter.finalizeRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1029:16: error: no member named 'cancelRootUpdate' in 'mlir::PatternRewriter' rewriter.cancelRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1108:14: error: no member named 'updateRootInPlace' in 'mlir::PatternRewriter' rewriter.updateRootInPlace(op->getParentOp(), [&]() { return; }); ~~~~~~~~ ^ 4 errors generated. Target @torch-mlir//:torch-mlir-opt failed to build ``` I'm still puzzled as to how this didn't fail with the CMake merge gating CI (do we not test Stablehlo builds/tests?). In any case, bumping our submodule to https://github.com/openxla/stablehlo/pull/1918 fixes it. It exposes a new failing lit test in TorchToStablehlo though, that I have looped stablehlo developers into ([here](https://discord.com/channels/999073994483433573/999074539138990131/1201235845391331419)). ``` bazel run @torch-mlir//test/Conversion:TorchToStablehlo/scatter.mlir.test ...external/torch-mlir/test/Conversion/TorchToStablehlo/scatter.mlir within split at <stdin>:1 offset :33:8: error: unexpected error: Expects non-empty reduction block for type inference %0 = torch.aten.scatter.src %arg0, %int0, %arg1, %arg2 : !torch.vtensor<[?,?],si64>, !torch.int, !torch.vtensor<[?,?],si64>, !torch.vtensor<[?,?],si64> -> !torch.vtensor<[?,?],si64> ^ LLVM ERROR: Failed to infer result type(s). ``` Bazel CI: https://github.com/sjain-stanford/torch-mlir/actions/runs/7732673480/job/21083102228
2024-02-01 06:21:17 +08:00
// CHECK: }) {padding = dense<0> : tensor<3x2xi64>, window_dilations = array<i64: 1, 1, 1>, window_dimensions = array<i64: 1, 3, 3>, window_strides = array<i64: 1, 2, 2>} : (tensor<?x?x?xf32>, tensor<?x?x?xi64>, tensor<f32>, tensor<i64>) -> (tensor<?x?x?xf32>, tensor<?x?x?xi64>)
build: update llvm tag to 74fb770d (#1539) * build: update llvm tag to 74fb770d This commit makes the following changes needed to update bump LLVM: + replace usages of `tensor::createPadScalarOp`, see https://reviews.llvm.org/D136493 + Update file checks
2022-11-01 15:27:09 +08:00
// CHECK: %[[T14:.*]] = torch_c.from_builtin_tensor %[[T13]]#0 : tensor<?x?x?xf32> -> !torch.vtensor<[?,?,?],f32>
// CHECK: %[[T15:.*]] = torch_c.from_builtin_tensor %[[T13]]#1 : tensor<?x?x?xi64> -> !torch.vtensor<[?,?,?],si64>
// CHECK: return %[[T14]], %[[T15]] : !torch.vtensor<[?,?,?],f32>, !torch.vtensor<[?,?,?],si64>
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
func.func @torch.aten.max_pool2d_with_indices(%arg0: !torch.vtensor<[?,?,?],f32>) -> (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[?,?,?],si64>) {
%int3 = torch.constant.int 3
%int2 = torch.constant.int 2
%false = torch.constant.bool false
%int0 = torch.constant.int 0
%int1 = torch.constant.int 1
%0 = torch.prim.ListConstruct %int3, %int3 : (!torch.int, !torch.int) -> !torch.list<int>
%1 = torch.prim.ListConstruct %int2, %int2 : (!torch.int, !torch.int) -> !torch.list<int>
%2 = torch.prim.ListConstruct %int0, %int0 : (!torch.int, !torch.int) -> !torch.list<int>
%3 = torch.prim.ListConstruct %int1, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
%result0, %result1 = torch.aten.max_pool2d_with_indices %arg0, %0, %1, %2, %3, %false : !torch.vtensor<[?,?,?],f32>, !torch.list<int>, !torch.list<int>, !torch.list<int>, !torch.list<int>, !torch.bool -> !torch.vtensor<[?,?,?],f32>, !torch.vtensor<[?,?,?],si64>
return %result0, %result1 : !torch.vtensor<[?,?,?],f32>, !torch.vtensor<[?,?,?],si64>
}
// -----
// CHECK-LABEL: func.func @torch.aten.avg_pool2d(
// CHECK-SAME: %[[VAL_0:.*]]: !torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[?,?,?,?],f32> {
// CHECK: %[[VAL_1:.*]] = torch_c.to_builtin_tensor %[[VAL_0]] : !torch.vtensor<[?,?,?,?],f32> -> tensor<?x?x?x?xf32>
// CHECK: %int3 = torch.constant.int 3
// CHECK: %int2 = torch.constant.int 2
// CHECK: %int1 = torch.constant.int 1
// CHECK: %false = torch.constant.bool false
// CHECK: %none = torch.constant.none
// CHECK: %[[VAL_2:.*]] = torch.prim.ListConstruct %int3, %int3 : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[VAL_3:.*]] = torch.prim.ListConstruct %int2, %int2 : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[VAL_4:.*]] = torch.prim.ListConstruct %int1, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[VAL_5:.*]] = stablehlo.constant dense<0.000000e+00> : tensor<f32>
// CHECK: %[[VAL_6:.*]] = "stablehlo.reduce_window"(%[[VAL_1]], %[[VAL_5]]) ({
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
// CHECK: ^bb0(%[[IVAL_0:.*]]: tensor<f32>, %[[IVAL_1:.*]]: tensor<f32>):
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[IVAL_2:.*]] = stablehlo.add %[[IVAL_0]], %[[IVAL_1]] : tensor<f32>
// CHECK: stablehlo.return %[[IVAL_2]] : tensor<f32>
Bump stablehlo to openxla/stablehlo@fd52182f76cadb82f2064fe5fc49a4fb4347a826 (#2821) With the recent LLVM integrate and changes from https://github.com/llvm/llvm-project/pull/78260, we hit this build error in Stablehlo (which is quite old). ``` external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1020:14: error: no member named 'startRootUpdate' in 'mlir::PatternRewriter' rewriter.startRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1026:16: error: no member named 'finalizeRootUpdate' in 'mlir::PatternRewriter' rewriter.finalizeRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1029:16: error: no member named 'cancelRootUpdate' in 'mlir::PatternRewriter' rewriter.cancelRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1108:14: error: no member named 'updateRootInPlace' in 'mlir::PatternRewriter' rewriter.updateRootInPlace(op->getParentOp(), [&]() { return; }); ~~~~~~~~ ^ 4 errors generated. Target @torch-mlir//:torch-mlir-opt failed to build ``` I'm still puzzled as to how this didn't fail with the CMake merge gating CI (do we not test Stablehlo builds/tests?). In any case, bumping our submodule to https://github.com/openxla/stablehlo/pull/1918 fixes it. It exposes a new failing lit test in TorchToStablehlo though, that I have looped stablehlo developers into ([here](https://discord.com/channels/999073994483433573/999074539138990131/1201235845391331419)). ``` bazel run @torch-mlir//test/Conversion:TorchToStablehlo/scatter.mlir.test ...external/torch-mlir/test/Conversion/TorchToStablehlo/scatter.mlir within split at <stdin>:1 offset :33:8: error: unexpected error: Expects non-empty reduction block for type inference %0 = torch.aten.scatter.src %arg0, %int0, %arg1, %arg2 : !torch.vtensor<[?,?],si64>, !torch.int, !torch.vtensor<[?,?],si64>, !torch.vtensor<[?,?],si64> -> !torch.vtensor<[?,?],si64> ^ LLVM ERROR: Failed to infer result type(s). ``` Bazel CI: https://github.com/sjain-stanford/torch-mlir/actions/runs/7732673480/job/21083102228
2024-02-01 06:21:17 +08:00
// CHECK{LITERAL}: }) {padding = dense<[[0, 0], [0, 0], [1, 1], [1, 1]]> : tensor<4x2xi64>, window_dilations = array<i64: 1, 1, 1, 1>, window_dimensions = array<i64: 1, 1, 3, 3>, window_strides = array<i64: 1, 1, 2, 2>} : (tensor<?x?x?x?xf32>, tensor<f32>) -> tensor<?x?x?x?xf32>
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[VAL_7:.*]] = stablehlo.constant dense<1.000000e+00> : tensor<f32>
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
// CHECK: %[[IDX_0:.*]] = arith.constant 0 : index
// CHECK: %[[VAL_8:.*]] = tensor.dim %[[VAL_1]], %[[IDX_0]] : tensor<?x?x?x?xf32>
// CHECK: %[[VAL_9:.*]] = arith.index_cast %[[VAL_8]] : index to i64
// CHECK: %[[IDX_1:.*]] = arith.constant 1 : index
// CHECK: %[[VAL_10:.*]] = tensor.dim %[[VAL_1]], %[[IDX_1]] : tensor<?x?x?x?xf32>
// CHECK: %[[VAL_11:.*]] = arith.index_cast %[[VAL_10]] : index to i64
// CHECK: %[[IDX_2:.*]] = arith.constant 2 : index
// CHECK: %[[VAL_12:.*]] = tensor.dim %[[VAL_1]], %[[IDX_2]] : tensor<?x?x?x?xf32>
// CHECK: %[[VAL_13:.*]] = arith.index_cast %[[VAL_12]] : index to i64
// CHECK: %[[IDX_3:.*]] = arith.constant 3 : index
// CHECK: %[[VAL_14:.*]] = tensor.dim %[[VAL_1]], %[[IDX_3]] : tensor<?x?x?x?xf32>
// CHECK: %[[VAL_15:.*]] = arith.index_cast %[[VAL_14]] : index to i64
// CHECK: %[[VAL_16:.*]] = tensor.from_elements %[[VAL_9]], %[[VAL_11]], %[[VAL_13]], %[[VAL_15]] : tensor<4xi64>
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[VAL_17:.*]] = stablehlo.dynamic_broadcast_in_dim %[[VAL_7]], %[[VAL_16]], dims = [] : (tensor<f32>, tensor<4xi64>) -> tensor<?x?x?x?xf32>
// CHECK: %[[VAL_18:.*]] = stablehlo.constant dense<0.000000e+00> : tensor<f32>
// CHECK: %[[VAL_19:.*]] = "stablehlo.reduce_window"(%[[VAL_17]], %[[VAL_18]]) ({
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
// CHECK: ^bb0(%[[IVAL_3:.*]]: tensor<f32>, %[[IVAL_4:.*]]: tensor<f32>):
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[IVAL_5:.*]] = stablehlo.add %[[IVAL_3]], %[[IVAL_4]] : tensor<f32>
// CHECK: stablehlo.return %[[IVAL_5]] : tensor<f32>
Bump stablehlo to openxla/stablehlo@fd52182f76cadb82f2064fe5fc49a4fb4347a826 (#2821) With the recent LLVM integrate and changes from https://github.com/llvm/llvm-project/pull/78260, we hit this build error in Stablehlo (which is quite old). ``` external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1020:14: error: no member named 'startRootUpdate' in 'mlir::PatternRewriter' rewriter.startRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1026:16: error: no member named 'finalizeRootUpdate' in 'mlir::PatternRewriter' rewriter.finalizeRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1029:16: error: no member named 'cancelRootUpdate' in 'mlir::PatternRewriter' rewriter.cancelRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1108:14: error: no member named 'updateRootInPlace' in 'mlir::PatternRewriter' rewriter.updateRootInPlace(op->getParentOp(), [&]() { return; }); ~~~~~~~~ ^ 4 errors generated. Target @torch-mlir//:torch-mlir-opt failed to build ``` I'm still puzzled as to how this didn't fail with the CMake merge gating CI (do we not test Stablehlo builds/tests?). In any case, bumping our submodule to https://github.com/openxla/stablehlo/pull/1918 fixes it. It exposes a new failing lit test in TorchToStablehlo though, that I have looped stablehlo developers into ([here](https://discord.com/channels/999073994483433573/999074539138990131/1201235845391331419)). ``` bazel run @torch-mlir//test/Conversion:TorchToStablehlo/scatter.mlir.test ...external/torch-mlir/test/Conversion/TorchToStablehlo/scatter.mlir within split at <stdin>:1 offset :33:8: error: unexpected error: Expects non-empty reduction block for type inference %0 = torch.aten.scatter.src %arg0, %int0, %arg1, %arg2 : !torch.vtensor<[?,?],si64>, !torch.int, !torch.vtensor<[?,?],si64>, !torch.vtensor<[?,?],si64> -> !torch.vtensor<[?,?],si64> ^ LLVM ERROR: Failed to infer result type(s). ``` Bazel CI: https://github.com/sjain-stanford/torch-mlir/actions/runs/7732673480/job/21083102228
2024-02-01 06:21:17 +08:00
// CHECK{LITERAL}: }) {padding = dense<[[0, 0], [0, 0], [1, 1], [1, 1]]> : tensor<4x2xi64>, window_dilations = array<i64: 1, 1, 1, 1>, window_dimensions = array<i64: 1, 1, 3, 3>, window_strides = array<i64: 1, 1, 2, 2>} : (tensor<?x?x?x?xf32>, tensor<f32>) -> tensor<?x?x?x?xf32>
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[VAL_20:.*]] = stablehlo.divide %[[VAL_6]], %[[VAL_19]] : tensor<?x?x?x?xf32>
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
// CHECK: %[[VAL_21:.*]] = torch_c.from_builtin_tensor %[[VAL_20]] : tensor<?x?x?x?xf32> -> !torch.vtensor<[?,?,?,?],f32>
// CHECK: return %[[VAL_21]] : !torch.vtensor<[?,?,?,?],f32>
func.func @torch.aten.avg_pool2d(%arg0: !torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[?,?,?,?],f32> {
%int3 = torch.constant.int 3
%int2 = torch.constant.int 2
%int1 = torch.constant.int 1
%false = torch.constant.bool false
%none = torch.constant.none
%0 = torch.prim.ListConstruct %int3, %int3 : (!torch.int, !torch.int) -> !torch.list<int>
%1 = torch.prim.ListConstruct %int2, %int2 : (!torch.int, !torch.int) -> !torch.list<int>
%2 = torch.prim.ListConstruct %int1, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
%3 = torch.aten.avg_pool2d %arg0, %0, %1, %2, %false, %false, %none : !torch.vtensor<[?,?,?,?],f32>, !torch.list<int>, !torch.list<int>, !torch.list<int>, !torch.bool, !torch.bool, !torch.none -> !torch.vtensor<[?,?,?,?],f32>
return %3 : !torch.vtensor<[?,?,?,?],f32>
}
// -----
build: update llvm tag to 74fb770d (#1539) * build: update llvm tag to 74fb770d This commit makes the following changes needed to update bump LLVM: + replace usages of `tensor::createPadScalarOp`, see https://reviews.llvm.org/D136493 + Update file checks
2022-11-01 15:27:09 +08:00
// CHECK-LABEL: func.func @torch.aten.avg_pool2d$count_include_pad(
// CHECK-SAME: %[[ARG0:.*]]: !torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[?,?,?,?],f32> {
// CHECK: %[[T0:.*]] = torch_c.to_builtin_tensor %[[ARG0]] : !torch.vtensor<[?,?,?,?],f32> -> tensor<?x?x?x?xf32>
// CHECK: %[[INT3:.*]] = torch.constant.int 3
// CHECK: %[[INT2:.*]] = torch.constant.int 2
// CHECK: %[[INT1:.*]] = torch.constant.int 1
// CHECK: %[[FALSE:.*]] = torch.constant.bool false
// CHECK: %[[TRUE:.*]] = torch.constant.bool true
// CHECK: %[[NONE:.*]] = torch.constant.none
// CHECK: %[[T1:.*]] = torch.prim.ListConstruct %[[INT3]], %[[INT3]] : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[T2:.*]] = torch.prim.ListConstruct %[[INT2]], %[[INT2]] : (!torch.int, !torch.int) -> !torch.list<int>
// CHECK: %[[T3:.*]] = torch.prim.ListConstruct %[[INT1]], %[[INT1]] : (!torch.int, !torch.int) -> !torch.list<int>
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[T4:.*]] = stablehlo.constant dense<0.000000e+00> : tensor<f32>
// CHECK: %[[T5:.*]] = "stablehlo.reduce_window"(%[[T0]], %[[T4]]) ({
build: update llvm tag to 74fb770d (#1539) * build: update llvm tag to 74fb770d This commit makes the following changes needed to update bump LLVM: + replace usages of `tensor::createPadScalarOp`, see https://reviews.llvm.org/D136493 + Update file checks
2022-11-01 15:27:09 +08:00
// CHECK: ^bb0(%[[ARG1:.*]]: tensor<f32>, %[[ARG2:.*]]: tensor<f32>):
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[T10:.*]] = stablehlo.add %[[ARG1]], %[[ARG2]] : tensor<f32>
// CHECK: stablehlo.return %[[T10]] : tensor<f32>
Bump stablehlo to openxla/stablehlo@fd52182f76cadb82f2064fe5fc49a4fb4347a826 (#2821) With the recent LLVM integrate and changes from https://github.com/llvm/llvm-project/pull/78260, we hit this build error in Stablehlo (which is quite old). ``` external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1020:14: error: no member named 'startRootUpdate' in 'mlir::PatternRewriter' rewriter.startRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1026:16: error: no member named 'finalizeRootUpdate' in 'mlir::PatternRewriter' rewriter.finalizeRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1029:16: error: no member named 'cancelRootUpdate' in 'mlir::PatternRewriter' rewriter.cancelRootUpdate(op); ~~~~~~~~ ^ external/stablehlo/stablehlo/transforms/StablehloRefineShapes.cpp:1108:14: error: no member named 'updateRootInPlace' in 'mlir::PatternRewriter' rewriter.updateRootInPlace(op->getParentOp(), [&]() { return; }); ~~~~~~~~ ^ 4 errors generated. Target @torch-mlir//:torch-mlir-opt failed to build ``` I'm still puzzled as to how this didn't fail with the CMake merge gating CI (do we not test Stablehlo builds/tests?). In any case, bumping our submodule to https://github.com/openxla/stablehlo/pull/1918 fixes it. It exposes a new failing lit test in TorchToStablehlo though, that I have looped stablehlo developers into ([here](https://discord.com/channels/999073994483433573/999074539138990131/1201235845391331419)). ``` bazel run @torch-mlir//test/Conversion:TorchToStablehlo/scatter.mlir.test ...external/torch-mlir/test/Conversion/TorchToStablehlo/scatter.mlir within split at <stdin>:1 offset :33:8: error: unexpected error: Expects non-empty reduction block for type inference %0 = torch.aten.scatter.src %arg0, %int0, %arg1, %arg2 : !torch.vtensor<[?,?],si64>, !torch.int, !torch.vtensor<[?,?],si64>, !torch.vtensor<[?,?],si64> -> !torch.vtensor<[?,?],si64> ^ LLVM ERROR: Failed to infer result type(s). ``` Bazel CI: https://github.com/sjain-stanford/torch-mlir/actions/runs/7732673480/job/21083102228
2024-02-01 06:21:17 +08:00
// CHECK{LITERAL}: }) {padding = dense<[[0, 0], [0, 0], [1, 1], [1, 1]]> : tensor<4x2xi64>, window_dilations = array<i64: 1, 1, 1, 1>, window_dimensions = array<i64: 1, 1, 3, 3>, window_strides = array<i64: 1, 1, 2, 2>} : (tensor<?x?x?x?xf32>, tensor<f32>) -> tensor<?x?x?x?xf32>
mhlo: migrate conversion to stablehlo (#1840) This patch replaces all MHLO operations with their StableHLO counterparts and adds a validation pass to ensure that no MHLO operations remain before translating all Stablehlo operations to the MHLO dialect for further lowering to the Linalg dialect. This patch also updates all lit tests so that they refer to the `convert-torch-to-stablehlo` pass and so that they check for StableHLO operations.
2023-02-02 21:29:47 +08:00
// CHECK: %[[T6:.*]] = stablehlo.constant dense<9> : tensor<i64>
// CHECK: %[[T7:.*]] = stablehlo.convert %[[T6]] : (tensor<i64>) -> tensor<f32>
build: update llvm tag to 74fb770d (#1539) * build: update llvm tag to 74fb770d This commit makes the following changes needed to update bump LLVM: + replace usages of `tensor::createPadScalarOp`, see https://reviews.llvm.org/D136493 + Update file checks
2022-11-01 15:27:09 +08:00
// CHECK: %[[T8:.*]] = chlo.broadcast_divide %[[T5]], %[[T7]] : (tensor<?x?x?x?xf32>, tensor<f32>) -> tensor<?x?x?x?xf32>
// CHECK: %[[T9:.*]] = torch_c.from_builtin_tensor %[[T8]] : tensor<?x?x?x?xf32> -> !torch.vtensor<[?,?,?,?],f32>
// CHECK: return %[[T9]] : !torch.vtensor<[?,?,?,?],f32>
[MHLO] Init MHLO pooling-like op conversion (#1141) * [MHLO] Init MHLO pooling-like op conversion and remove 'op' suffix in filenames Co-authored-by: Bairen Yi <yibairen.byron@bytedance.com> Co-authored-by: Jiawei Wu <xremold@gmail.com> Co-authored-by: Tianyou Guo tianyou.gty@alibaba-inc.com Co-authored-by: Xu Yan <yancey.yx@alibaba-inc.com> Co-authored-by: Ziheng Jiang <ziheng.jiang@bytedance.com> See RFC #999
2022-08-04 12:34:22 +08:00
func.func @torch.aten.avg_pool2d$count_include_pad(%arg0: !torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[?,?,?,?],f32> {
%int3 = torch.constant.int 3
%int2 = torch.constant.int 2
%int1 = torch.constant.int 1
%false = torch.constant.bool false
%true = torch.constant.bool true
%none = torch.constant.none
%0 = torch.prim.ListConstruct %int3, %int3 : (!torch.int, !torch.int) -> !torch.list<int>
%1 = torch.prim.ListConstruct %int2, %int2 : (!torch.int, !torch.int) -> !torch.list<int>
%2 = torch.prim.ListConstruct %int1, %int1 : (!torch.int, !torch.int) -> !torch.list<int>
%3 = torch.aten.avg_pool2d %arg0, %0, %1, %2, %false, %true, %none : !torch.vtensor<[?,?,?,?],f32>, !torch.list<int>, !torch.list<int>, !torch.list<int>, !torch.bool, !torch.bool, !torch.none -> !torch.vtensor<[?,?,?,?],f32>
return %3 : !torch.vtensor<[?,?,?,?],f32>
llvm: update tag to 061e0189 (#1180) Summary of changes: - Switch to C++17 (similar to https://reviews.llvm.org/D131348) - Update MHLO to build with LLVM commit hash 061e0189 - Replace deprecated `hasValue()` and `getValue()` with `has_value()` and `value()` respectively (https://reviews.llvm.org/D131349) - Use `TypedAttr` (https://reviews.llvm.org/D130092) - Use updated assembly format of `mhlo.compare` op (commit d03ef01e70fbf9afd0fa1976fbb7ed31838929b3 in MHLO repo)
2022-08-09 11:17:35 +08:00
}