[TORCH] Add support for aten.hinge_embedding_loss Op

include/torch-mlir/Dialect/Torch/IR/GeneratedTorchOps.td

@@ -9478,6 +9478,32 @@ def Torch_AtenNllLossBackwardOp : Torch_Op<"aten.nll_loss_backward", [
   }];
 }
 
+def Torch_AtenHingeEmbeddingLossOp : Torch_Op<"aten.hinge_embedding_loss", [
+    AllowsTypeRefinement,
+    HasValueSemantics,
+    ReadOnly
+  ]> {
+  let summary = "Generated op for `aten::hinge_embedding_loss : (Tensor, Tensor, float, int) -> (Tensor)`";
+  let arguments = (ins
+    AnyTorchTensorType:$self,
+    AnyTorchTensorType:$target,
+    Torch_FloatType:$margin,
+    Torch_IntType:$reduction
+  );
+  let results = (outs
+    AnyTorchOptionalTensorType:$result
+  );
+  let hasCustomAssemblyFormat = 1;
+  let extraClassDefinition = [{
+    ParseResult AtenHingeEmbeddingLossOp::parse(OpAsmParser &parser, OperationState &result) {
+      return parseDefaultTorchOp(parser, result, 4, 1);
+    }
+    void AtenHingeEmbeddingLossOp::print(OpAsmPrinter &printer) {
+      printDefaultTorchOp(printer, *this, 4, 1);
+    }
+  }];
+}
+
 def Torch_AtenPoissonNllLossOp : Torch_Op<"aten.poisson_nll_loss", [
     AllowsTypeRefinement,
     HasValueSemantics,
@@ -9496,7 +9522,7 @@ def Torch_AtenPoissonNllLossOp : Torch_Op<"aten.poisson_nll_loss", [
     AnyTorchOptionalTensorType:$result
   );
   let hasCustomAssemblyFormat = 1;
-  let extraClassDefinition = [{
+  let extraClassDefinition = [{   
     ParseResult AtenPoissonNllLossOp::parse(OpAsmParser &parser, OperationState &result) {
       return parseDefaultTorchOp(parser, result, 6, 1);
     }

lib/Dialect/Torch/Transforms/AbstractInterpLibrary.cpp

@@ -10729,6 +10729,20 @@ StringRef mlir::torch::Torch::getAbstractInterpLibrary() {
 "    %0 = call @__torch__.torch.jit._shape_functions.unary(%arg1) : (!torch.list<int>) -> !torch.list<int>\n"
 "    return %0 : !torch.list<int>\n"
 "  }\n"
+"  func.func @\"__torch_mlir_shape_fn.aten.hinge_embedding_loss\"(%arg0: !torch.list<int>, %arg1: !torch.list<int>, %arg2: !torch.float, %arg3: !torch.int) -> !torch.list<int> {\n"
+"    %int1 = torch.constant.int 1\n"
+"    %int2 = torch.constant.int 2\n"
+"    %0 = torch.prim.ListConstruct %int1, %int2 : (!torch.int, !torch.int) -> !torch.list<int>\n"
+"    %1 = torch.aten.__contains__.int_list %0, %arg3 : !torch.list<int>, !torch.int -> !torch.bool\n"
+"    %2 = torch.prim.If %1 -> (!torch.list<int>) {\n"
+"      %3 = torch.prim.ListConstruct  : () -> !torch.list<int>\n"
+"      torch.prim.If.yield %3 : !torch.list<int>\n"
+"    } else {\n"
+"      %3 = func.call @__torch__.torch.jit._shape_functions.unary(%arg0) : (!torch.list<int>) -> !torch.list<int>\n"
+"      torch.prim.If.yield %3 : !torch.list<int>\n"
+"    }\n"
+"    return %2 : !torch.list<int>\n"
+"  }\n"
 "  func.func @\"__torch_mlir_shape_fn.aten.mse_loss\"(%arg0: !torch.list<int>, %arg1: !torch.list<int>, %arg2: !torch.int) -> !torch.list<int> {\n"
 "    %int0 = torch.constant.int 0\n"
 "    %0 = torch.aten.eq.int %arg2, %int0 : !torch.int, !torch.int -> !torch.bool\n"
@@ -13427,6 +13441,14 @@ StringRef mlir::torch::Torch::getAbstractInterpLibrary() {
 "    }\n"
 "    return %6 : !torch.int\n"
 "  }\n"
+"  func.func @\"__torch_mlir_dtype_fn.aten.hinge_embedding_loss\"(%arg0: !torch.tuple<int, int>, %arg1: !torch.tuple<int, int>, %arg2: !torch.float, %arg3: !torch.int) -> !torch.int {\n"
+"    %0:2 = torch.prim.TupleUnpack %arg0 : !torch.tuple<int, int> -> !torch.int, !torch.int\n"
+"    %1:2 = torch.prim.TupleUnpack %arg1 : !torch.tuple<int, int> -> !torch.int, !torch.int\n"
+"    %2 = torch.prim.ListConstruct %0#0, %1#0 : (!torch.int, !torch.int) -> !torch.list<optional<int>>\n"
+"    %3 = torch.prim.ListConstruct %0#1, %1#1 : (!torch.int, !torch.int) -> !torch.list<int>\n"
+"    %4 = call @__torch__.torch_mlir.jit_ir_importer.build_tools.library_generator.promote_dtypes(%2, %3) : (!torch.list<optional<int>>, !torch.list<int>) -> !torch.int\n"
+"    return %4 : !torch.int\n"
+"  }\n"
 "  func.func @\"__torch_mlir_dtype_fn.aten.max_pool2d_with_indices_backward\"(%arg0: !torch.tuple<int, int>, %arg1: !torch.tuple<int, int>, %arg2: !torch.list<int>, %arg3: !torch.list<int>, %arg4: !torch.list<int>, %arg5: !torch.list<int>, %arg6: !torch.bool, %arg7: !torch.tuple<int, int>) -> !torch.int {\n"
 "    %none = torch.constant.none\n"
 "    %str = torch.constant.str \"AssertionError: \"\n"

lib/Dialect/Torch/Transforms/DecomposeComplexOps.cpp

@@ -10553,6 +10553,88 @@ class DecomposeAtenNllLossForwardOp
 };
 } // namespace
 
+namespace {
+// Decompostion of aten.hinge_embedding_loss op
+// Ref:
+// https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/native/Loss.cpp#L182
+// The Hinge Embedding Loss:
+//           | input,                  if target == 1
+// loss(x) = |
+//           | max(0, margin - input), if target == -1
+// target tensor may have values other than 1 and -1
+class DecomposeHingeEmbeddingLoss
+    : public OpRewritePattern<AtenHingeEmbeddingLossOp> {
+  using OpRewritePattern<AtenHingeEmbeddingLossOp>::OpRewritePattern;
+  LogicalResult matchAndRewrite(AtenHingeEmbeddingLossOp op,
+                                PatternRewriter &rewriter) const override {
+    Location loc = op.getLoc();
+    auto input = op.getSelf();
+    auto target = op.getTarget();
+
+    auto inputTy = dyn_cast<ValueTensorType>(input.getType());
+    if (!inputTy.hasDtype() || !inputTy.hasSizes())
+      return rewriter.notifyMatchFailure(op, "input must have dtype and size");
+
+    auto targetTy = dyn_cast<ValueTensorType>(target.getType());
+    if (!targetTy.hasDtype() || !targetTy.hasSizes())
+      return rewriter.notifyMatchFailure(op, "target must have dtype and size");
+
+    int64_t reduction;
+    if (!matchPattern(op.getReduction(), m_TorchConstantInt(&reduction))) {
+      return rewriter.notifyMatchFailure(op,
+                                         "reduction should be a constant int!");
+    }
+    auto resultTy = dyn_cast<ValueTensorType>(op.getType());
+    Value minusOne = getConstantWithGivenDtypeAndValue(rewriter, loc, -1,
+                                                       targetTy.getDtype());
+    Value one = getConstantWithGivenDtypeAndValue(rewriter, loc, 1,
+                                                  targetTy.getDtype());
+    Value zero = getConstantWithGivenDtypeAndValue(rewriter, loc, 0,
+                                                   targetTy.getDtype());
+    Value alpha =
+        rewriter.create<ConstantIntOp>(loc, rewriter.getI64IntegerAttr(1));
+    auto boolType = targetTy.getWithSizesAndDtype(targetTy.getSizes(),
+                                                  rewriter.getI1Type());
+    // input - margin
+    auto inputMinusMargin = rewriter.create<AtenSubScalarOp>(
+        loc, inputTy, input, op.getMargin(), alpha);
+    // multiply by -1 to get margin - input
+    auto marginDiff = rewriter.create<AtenMulScalarOp>(
+        loc, inputTy, inputMinusMargin, minusOne);
+    // max(0, margin - input) => clamping the minimum value of margin - input at
+    // 0
+    auto marginClamp =
+        rewriter.create<AtenClampMinOp>(loc, inputTy, marginDiff, zero);
+    // Compute mask: target != 1
+    auto targetNotOne =
+        rewriter.create<AtenNeScalarOp>(loc, boolType, target, one);
+    // If target != 1 use marginClamp otherwise 0.
+    auto outputMargin = rewriter.create<AtenWhereScalarOtherOp>(
+        loc, inputTy, targetNotOne, marginClamp, zero);
+    // Compute mask: target != -1
+    auto targetNotMinusOne =
+        rewriter.create<AtenNeScalarOp>(loc, boolType, target, minusOne);
+    // If target != -1 use the original input. Otherwise 0.
+    auto outputSelf = rewriter.create<AtenWhereScalarOtherOp>(
+        loc, inputTy, targetNotMinusOne, input, zero);
+    // Add : outputMargin + outputSelf
+    auto output = rewriter.create<AtenAddTensorOp>(loc, inputTy, outputMargin,
+                                                   outputSelf, /*alpha=*/alpha);
+    Value loss = output;
+    Value none = rewriter.create<ConstantNoneOp>(loc);
+    // reduction: mean
+    if (reduction == 1) {
+      loss = rewriter.create<AtenMeanOp>(loc, resultTy, output, none);
+    } else if (reduction == 2) {
+      // reduction: sum
+      loss = rewriter.create<AtenSumOp>(loc, resultTy, output, none);
+    }
+    rewriter.replaceOp(op, loss);
+    return success();
+  }
+};
+} // namespace
+
 namespace {
 class DecomposeAtenPoissonNllLossOp
     : public OpRewritePattern<AtenPoissonNllLossOp> {
@@ -12692,6 +12774,7 @@ class DecomposeComplexOpsPass
     addPatternIfTargetOpIsIllegal<DecomposeAtenOneHotOp>(patterns);
     addPatternIfTargetOpIsIllegal<DecomposeAtenCrossEntropyLossOp>(patterns);
     addPatternIfTargetOpIsIllegal<DecomposeAtenNllLossForwardOp>(patterns);
+    addPatternIfTargetOpIsIllegal<DecomposeHingeEmbeddingLoss>(patterns);
     addPatternIfTargetOpIsIllegal<DecomposeAtenPoissonNllLossOp>(patterns);
     addPatternIfTargetOpIsIllegal<DecomposeAtenBinaryCrossEntropyWithLogitsOp>(
         patterns);

lib/Dialect/Torch/Transforms/LowerToBackendContract.cpp

@@ -539,6 +539,7 @@ static void markDecomposedOpsAsIllegal(MLIRContext *context,
   target.addIllegalOp<AtenLerpTensorOp>();
   target.addIllegalOp<AtenMseLossOp>();
   target.addIllegalOp<AtenL1LossOp>();
+  target.addIllegalOp<AtenHingeEmbeddingLossOp>();
   target.addIllegalOp<AtenPoissonNllLossOp>();
   target.addIllegalOp<AtenRandintLowOp>();
   target.addIllegalOp<AtenRandintOp>();

projects/pt1/e2e_testing/xfail_sets.py

@@ -1804,6 +1804,9 @@
     "L1LossMeanReductionModule_basic",
     "L1LossNoReductionModule_basic",
     "L1LossSumReductionModule_basic",
+    "HingeEmbeddingLossReductionMeanModule_basic",
+    "HingeEmbeddingLossReductionSumModule_basic",
+    "HingeEmbeddingLossReductionNoneModule_basic",
     "PixelShuffleModuleStaticRank3Int64_basic",
     "PixelShuffleModuleStaticRank4Float32_basic",
     "RandIntLowModule_basic",
@@ -2978,6 +2981,10 @@
     "GtFloatIntModule_basic",
     "GtIntModule_basic",
     "HardtanhBackward_basic",
+    "HingeEmbeddingLossBasicModule_basic",
+    "HingeEmbeddingLossReductionMeanModule_basic",
+    "HingeEmbeddingLossReductionSumModule_basic",
+    "HingeEmbeddingLossReductionNoneModule_basic",
     "ElementwiseHeavisideModule_basic",
     "ElementwiseHeavisideIntModule_basic",
     "ElementwiseHeavisideNoBroadcastModule_basic",
@@ -3996,6 +4003,10 @@
     "NllLossStaticModule_mean_basic",
     "NllLossStaticModule_sum_basic",
     "NllLossStaticModule_weight_basic",
+    "HingeEmbeddingLossBasicModule_basic",
+    "HingeEmbeddingLossReductionMeanModule_basic",
+    "HingeEmbeddingLossReductionSumModule_basic",
+    "HingeEmbeddingLossReductionNoneModule_basic",
     "KlDivLossModule_default_basic",
     "KlDivLossModule_reduction_is_none_basic",
     "KlDivLossModule_reduction_is_none_log_target_is_true_basic",

projects/pt1/python/torch_mlir/jit_ir_importer/build_tools/abstract_interp_lib_gen.py

@@ -2197,6 +2197,11 @@ def aten〇nll_loss_forward〡shape(self: List[int], target: List[int], weight:
 def aten〇nll_loss_backward〡shape(grad_output: List[int], self: List[int], target: List[int], weight: Optional[List[int]], reduction: int, ignore_index: int, total_weight: List[int]) -> List[int]:
     return upstream_shape_functions.unary(self)
 
+def aten〇hinge_embedding_loss〡shape(self: List[int], target: List[int], margin: float = 1., reduction: int = 1) -> List[int]:
+    if reduction in [1,2]:
+        return []
+    return upstream_shape_functions.unary(self)
+
 # TODO: upstream this
 def aten〇mse_loss〡shape(self: List[int], target: List[int], reduction: int = 1) -> List[int]:
     if reduction == 0:
@@ -3993,6 +3998,13 @@ def aten〇nll_loss_backward〡dtype(grad_output_rank_dtype: Tuple[int, int], se
         return torch.int64
     return result
 
+def aten〇hinge_embedding_loss〡dtype(self_rank_dtype: Tuple[int, int], target_rank_dtype: Tuple[int, int], margin: float = 1., reduction: int = 1) -> int:
+    self_rank, self_dtype = self_rank_dtype
+    target_rank, target_dtype = target_rank_dtype
+    ranks: List[Optional[int]] = [self_rank, target_rank]
+    dtypes = [self_dtype, target_dtype]
+    return promote_dtypes(ranks, dtypes)
+
 @check_dtype_function(_check_tensors_with_the_same_dtype(
     None, [(2, 4, 7, 6), (2, 4, 6, 5)], None, None,
     [2, 2], [1, 1], [1, 1], [1, 1], False, TensorOfShape(2, 4, 7, 6, dtype=torch.int64)) +

projects/pt1/python/torch_mlir/jit_ir_importer/build_tools/torch_ods_gen.py

@@ -761,6 +761,7 @@ def emit_with_mutating_variants(key, **kwargs):
     emit(
         "aten::nll_loss_backward : (Tensor, Tensor, Tensor, Tensor?, int, int, Tensor) -> (Tensor)"
     )
+    emit("aten::hinge_embedding_loss : (Tensor, Tensor, float, int) -> (Tensor)")
     emit(
         "aten::poisson_nll_loss : (Tensor, Tensor, bool, bool, float, int) -> (Tensor)"
     )

projects/pt1/python/torch_mlir_e2e_test/test_suite/reduction.py

@@ -2455,6 +2455,95 @@ def BinaryCrossEntropyWithLogitsStaticModule_basic(module, tu: TestUtils):
 # ==============================================================================
 
 
+class HingeEmbeddingLossBasicModule(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    @export
+    @annotate_args(
+        [
+            None,
+            ([-1, -1, -1], torch.float32, True),
+            ([-1, -1, -1], torch.float32, True),
+        ]
+    )
+    def forward(self, input, target):
+        return torch.ops.aten.hinge_embedding_loss(
+            input, target, margin=1.5, reduction=1
+        )
+
+
+@register_test_case(module_factory=lambda: HingeEmbeddingLossBasicModule())
+def HingeEmbeddingLossBasicModule_basic(module, tu: TestUtils):
+    module.forward(tu.rand(1, 2, 3), tu.rand(1, 2, 3))
+
+
+class HingeEmbeddingLossReductionMeanModule(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    @export
+    @annotate_args(
+        [
+            None,
+            ([-1, -1], torch.float32, True),
+            ([-1, -1], torch.float32, True),
+        ]
+    )
+    def forward(self, input, target):
+        return torch.ops.aten.hinge_embedding_loss(input, target, reduction=1)
+
+
+@register_test_case(module_factory=lambda: HingeEmbeddingLossReductionMeanModule())
+def HingeEmbeddingLossReductionMeanModule_basic(module, tu: TestUtils):
+    module.forward(tu.rand(8, 1), tu.rand(1, 1))
+
+
+class HingeEmbeddingLossReductionSumModule(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    @export
+    @annotate_args(
+        [
+            None,
+            ([-1, -1], torch.float32, True),
+            ([-1, -1], torch.float32, True),
+        ]
+    )
+    def forward(self, input, target):
+        return torch.ops.aten.hinge_embedding_loss(input, target, reduction=2)
+
+
+@register_test_case(module_factory=lambda: HingeEmbeddingLossReductionSumModule())
+def HingeEmbeddingLossReductionSumModule_basic(module, tu: TestUtils):
+    module.forward(tu.rand(2, 5), tu.rand(1, 1))
+
+
+class HingeEmbeddingLossReductionNoneModule(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    @export
+    @annotate_args(
+        [
+            None,
+            ([-1, -1], torch.float32, True),
+            ([-1, -1], torch.float32, True),
+        ]
+    )
+    def forward(self, input, target):
+        return torch.ops.aten.hinge_embedding_loss(input, target, margin=1.0)
+
+
+@register_test_case(module_factory=lambda: HingeEmbeddingLossReductionNoneModule())
+def HingeEmbeddingLossReductionNoneModule_basic(module, tu: TestUtils):
+    module.forward(tu.rand(8, 5), tu.rand(1, 1))
+
+
+# ==============================================================================
+
+
 class TraceModule(torch.nn.Module):
     def __init__(self) -> None:
         super().__init__()