Implement lowering of aten.cosh op. (#2635)

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

@@ -64,6 +64,51 @@ def Torch_AtenTanh_Op : Torch_Op<"aten.tanh_", [
   }];
 }
 
+def Torch_AtenCoshOp : Torch_Op<"aten.cosh", [
+    AllowsTypeRefinement,
+    HasValueSemantics,
+    ReadOnly
+  ]> {
+  let summary = "Generated op for `aten::cosh : (Tensor) -> (Tensor)`";
+  let arguments = (ins
+    AnyTorchTensorType:$self
+  );
+  let results = (outs
+    AnyTorchTensorType:$result
+  );
+  let hasCustomAssemblyFormat = 1;
+  let extraClassDefinition = [{
+    ParseResult AtenCoshOp::parse(OpAsmParser &parser, OperationState &result) {
+      return parseDefaultTorchOp(parser, result, 1, 1);
+    }
+    void AtenCoshOp::print(OpAsmPrinter &printer) {
+      printDefaultTorchOp(printer, *this, 1, 1);
+    }
+  }];
+}
+
+def Torch_AtenCosh_Op : Torch_Op<"aten.cosh_", [
+    IsTrailingUnderscoreInplaceVariant,
+    AllowsTypeRefinement
+  ]> {
+  let summary = "Generated op for `aten::cosh_ : (Tensor) -> (Tensor)`";
+  let arguments = (ins
+    Torch_NonValueTensorType:$self
+  );
+  let results = (outs
+    Torch_NonValueTensorType:$result
+  );
+  let hasCustomAssemblyFormat = 1;
+  let extraClassDefinition = [{
+    ParseResult AtenCosh_Op::parse(OpAsmParser &parser, OperationState &result) {
+      return parseDefaultTorchOp(parser, result, 1, 1);
+    }
+    void AtenCosh_Op::print(OpAsmPrinter &printer) {
+      printDefaultTorchOp(printer, *this, 1, 1);
+    }
+  }];
+}
+
 def Torch_AtenHardtanhOp : Torch_Op<"aten.hardtanh", [
     AllowsTypeRefinement,
     HasValueSemantics,

lib/Conversion/TorchToLinalg/Uncategorized.cpp

@@ -220,6 +220,10 @@ static Value createLinalgPayloadCalculationForElementwiseOp(
     return createCalculationForMathOpWithDtypeConversion<math::TanhOp>(
         b, converter, payloadArgs[0], op);
   }
+  if (isa<AtenCoshOp>(op)) {
+    return createCalculationForMathOpWithDtypeConversion<math::CoshOp>(
+        b, converter, payloadArgs[0], op);
+  }
   if (isa<AtenExpOp>(op)) {
     return createCalculationForMathOpWithDtypeConversion<math::ExpOp>(
         b, converter, payloadArgs[0], op);
@@ -1311,7 +1315,7 @@ class ConvertElementwiseOp : public ConversionPattern {
   LogicalResult
   matchAndRewrite(Operation *op, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const override {
-    if (!isa<AtenTanhOp, AtenReluOp, AtenPreluOp, AtenGeluOp,
+    if (!isa<AtenTanhOp, AtenCoshOp, AtenReluOp, AtenPreluOp, AtenGeluOp,
              AtenGeluBackwardOp, AtenAddTensorOp, AtenMulTensorOp,
              AtenDivTensorOp, AtenDivTensorModeOp, AtenSubTensorOp, AtenAtan2Op,
              AtenLerpTensorOp, AtenSigmoidOp, AtenExpOp, AtenExpm1Op,
@@ -1964,7 +1968,7 @@ void mlir::torch::torch_to_linalg::populateUncategorizedPatternsAndLegality(
     ConversionTarget &target) {
   MLIRContext *context = patterns.getContext();
   target.addIllegalOp<
-      AtenTanhOp, AtenReluOp, AtenGeluOp, AtenGeluBackwardOp, AtenAddTensorOp,
+      AtenTanhOp, AtenCoshOp, AtenReluOp, AtenGeluOp, AtenGeluBackwardOp, AtenAddTensorOp,
       AtenMulTensorOp, AtenDivTensorOp, AtenDivTensorModeOp, AtenSubTensorOp,
       AtenLerpTensorOp, AtenSigmoidOp, AtenMinimumOp, AtenAtan2Op,
       AtenMaximumOp, AtenToDtypeOp, AtenClampOp, AtenClampTensorOp,

lib/Dialect/Torch/Transforms/AbstractInterpLibrary.cpp

@@ -6242,6 +6242,10 @@ StringRef mlir::torch::Torch::getAbstractInterpLibrary() {
 "    %0 = call @__torch__.torch.jit._shape_functions.unary(%arg0) : (!torch.list<int>) -> !torch.list<int>\n"
 "    return %0 : !torch.list<int>\n"
 "  }\n"
+"  func.func @\"__torch_mlir_shape_fn.aten.cosh\"(%arg0: !torch.list<int>) -> !torch.list<int> {\n"
+"    %0 = call @__torch__.torch.jit._shape_functions.unary(%arg0) : (!torch.list<int>) -> !torch.list<int>\n"
+"    return %0 : !torch.list<int>\n"
+"  }\n"
 "  func.func @\"__torch_mlir_shape_fn.aten.tanh\"(%arg0: !torch.list<int>) -> !torch.list<int> {\n"
 "    %0 = call @__torch__.torch.jit._shape_functions.unary(%arg0) : (!torch.list<int>) -> !torch.list<int>\n"
 "    return %0 : !torch.list<int>\n"
@@ -8523,7 +8527,7 @@ StringRef mlir::torch::Torch::getAbstractInterpLibrary() {
 "    %0:2 = torch.prim.TupleUnpack %arg0 : !torch.tuple<int, int> -> !torch.int, !torch.int\n"
 "    return %0#1 : !torch.int\n"
 "  }\n"
-"  func.func @\"__torch_mlir_dtype_fn.aten.tanh\"(%arg0: !torch.tuple<int, int>) -> !torch.int {\n"
+"  func.func @\"__torch_mlir_dtype_fn.aten.cosh\"(%arg0: !torch.tuple<int, int>) -> !torch.int {\n"
 "    %0:2 = torch.prim.TupleUnpack %arg0 : !torch.tuple<int, int> -> !torch.int, !torch.int\n"
 "    %1 = call @__torch__._get_dtype_of_floating_point_op(%0#1) : (!torch.int) -> !torch.int\n"
 "    return %1 : !torch.int\n"
@@ -8565,6 +8569,11 @@ StringRef mlir::torch::Torch::getAbstractInterpLibrary() {
 "    %0 = torch.prim.ListConstruct %int5, %int15, %int6, %int7 : (!torch.int, !torch.int, !torch.int, !torch.int) -> !torch.list<int>\n"
 "    return %0 : !torch.list<int>\n"
 "  }\n"
+"  func.func @\"__torch_mlir_dtype_fn.aten.tanh\"(%arg0: !torch.tuple<int, int>) -> !torch.int {\n"
+"    %0:2 = torch.prim.TupleUnpack %arg0 : !torch.tuple<int, int> -> !torch.int, !torch.int\n"
+"    %1 = call @__torch__._get_dtype_of_floating_point_op(%0#1) : (!torch.int) -> !torch.int\n"
+"    return %1 : !torch.int\n"
+"  }\n"
 "  func.func @\"__torch_mlir_dtype_fn.aten.exp\"(%arg0: !torch.tuple<int, int>) -> !torch.int {\n"
 "    %0:2 = torch.prim.TupleUnpack %arg0 : !torch.tuple<int, int> -> !torch.int, !torch.int\n"
 "    %1 = call @__torch__._get_dtype_of_floating_point_op(%0#1) : (!torch.int) -> !torch.int\n"

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

@@ -62,6 +62,9 @@ def aten〇tril〡shape(self: List[int], diagonal: int = 0) -> List[int]:
 def aten〇atan〡shape(self: List[int]) -> List[int]:
     return upstream_shape_functions.unary(self)
 
+def aten〇cosh〡shape(self: List[int]) -> List[int]:
+    return upstream_shape_functions.unary(self)
+
 def aten〇tanh〡shape(self: List[int]) -> List[int]:
     return upstream_shape_functions.unary(self)
 
@@ -1538,6 +1541,13 @@ def prims〇split_dim〡dtype(a_rank_dtype: Tuple[int, int], dim: int, outer_len
     _, a_dtype = a_rank_dtype
     return a_dtype
 
+
+@check_dtype_function(_check_tensors_with_the_same_dtype(num_of_tensors=1))
+def aten〇cosh〡dtype(self_rank_dtype: Tuple[int, int]) -> int:
+    self_rank, self_dtype = self_rank_dtype
+    return _get_dtype_of_floating_point_op(self_dtype)
+
+
 @check_dtype_function(_check_tensors_with_the_same_dtype(num_of_tensors=1))
 def aten〇tanh〡dtype(self_rank_dtype: Tuple[int, int]) -> int:
     self_rank, self_dtype = self_rank_dtype

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

@@ -256,6 +256,7 @@ def emit_with_mutating_variants(key, **kwargs):
     # Elementwise tensor compute ops
     for key in [
             "aten::tanh : (Tensor) -> (Tensor)",
+            "aten::cosh : (Tensor) -> (Tensor)",
             "aten::hardtanh : (Tensor, Scalar, Scalar) -> (Tensor)",
             "aten::elu : (Tensor, Scalar, Scalar, Scalar) -> (Tensor)",
             "aten::relu : (Tensor) -> (Tensor)",

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

@@ -63,6 +63,50 @@ def ElementwiseUnaryIntModule_basic(module, tu: TestUtils):
 # ==============================================================================
 
 
+class ElementwiseCoshModule(torch.nn.Module):
+
+    def __init__(self):
+        super().__init__()
+
+    @export
+    @annotate_args([
+        None,
+        ([-1, -1], torch.float32, True),
+    ])
+    def forward(self, a):
+        return torch.cosh(a)
+
+
+@register_test_case(module_factory=lambda: ElementwiseCoshModule())
+def ElementwiseCoshModule_basic(module, tu: TestUtils):
+    module.forward(tu.rand(3, 4))
+
+
+# ==============================================================================
+
+
+class ElementwiseCoshIntModule(torch.nn.Module):
+
+    def __init__(self):
+        super().__init__()
+
+    @export
+    @annotate_args([
+        None,
+        ([-1, -1], torch.int32, True),
+    ])
+    def forward(self, a):
+        return torch.cosh(a)
+
+
+@register_test_case(module_factory=lambda: ElementwiseCoshIntModule())
+def ElementwiseCoshIntModule_basic(module, tu: TestUtils):
+    module.forward(tu.randint(3, 4, low=1, high=10).to(torch.int32))
+
+
+# ==============================================================================
+
+
 class ElementwiseBinaryModule(torch.nn.Module):
 
     def __init__(self):