[Torch] : Implement lowering of torch.frac op

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

@@ -5482,6 +5482,29 @@ def Torch_AtenPolarOp : Torch_Op<"aten.polar", [
   }];
 }
 
+def Torch_AtenFracOp : Torch_Op<"aten.frac", [
+    AllowsTypeRefinement,
+    HasValueSemantics,
+    ReadOnly
+  ]> {
+  let summary = "Generated op for `aten::frac : (Tensor) -> (Tensor)`";
+  let arguments = (ins
+    AnyTorchTensorType:$self
+  );
+  let results = (outs
+    AnyTorchOptionalTensorType:$result
+  );
+  let hasCustomAssemblyFormat = 1;
+  let extraClassDefinition = [{
+    ParseResult AtenFracOp::parse(OpAsmParser &parser, OperationState &result) {
+      return parseDefaultTorchOp(parser, result, 1, 1);
+    }
+    void AtenFracOp::print(OpAsmPrinter &printer) {
+      printDefaultTorchOp(printer, *this, 1, 1);
+    }
+  }];
+}
+
 def Torch_AtenUnbindCopyIntOp : Torch_Op<"aten.unbind_copy.int", [
     AllowsTypeRefinement,
     HasValueSemantics,

lib/Dialect/Torch/Transforms/AbstractInterpLibrary.cpp

@@ -6703,6 +6703,10 @@ 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.frac\"(%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.log\"(%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"
@@ -11853,6 +11857,10 @@ 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.frac\"(%arg0: !torch.tuple<int, int>) -> !torch.int {\n"
+"    %int6 = torch.constant.int 6\n"
+"    return %int6 : !torch.int\n"
+"  }\n"
 "  func.func @\"__torch_mlir_dtype_fn.aten.clamp_max\"(%arg0: !torch.tuple<int, int>, %arg1: !torch.number) -> !torch.int {\n"
 "    %int4 = torch.constant.int 4\n"
 "    %int11 = torch.constant.int 11\n"

lib/Dialect/Torch/Transforms/DecomposeComplexOps.cpp

@@ -8012,6 +8012,34 @@ class DecomposeAtenTruncOp : public OpRewritePattern<AtenTruncOp> {
 };
 } // namespace
 
+namespace {
+// decompose `frac(x)` to `x - sign(x) * floor(abs(x))`
+class DecomposeAtenFracOp : public OpRewritePattern<AtenFracOp> {
+  using OpRewritePattern::OpRewritePattern;
+  LogicalResult matchAndRewrite(AtenFracOp op,
+                                PatternRewriter &rewriter) const override {
+    Location loc = op.getLoc();
+    Value self = op.getSelf();
+
+    auto resultTy = dyn_cast<ValueTensorType>(op.getType());
+    if (!resultTy || !resultTy.hasDtype()) {
+      return rewriter.notifyMatchFailure(op, "result must have dtype");
+    }
+
+    if (isa<mlir::FloatType>(resultTy.getDtype())) {
+      Value trunc = rewriter.create<AtenTruncOp>(loc, resultTy, self);
+      Value alpha =
+          rewriter.create<ConstantFloatOp>(loc, rewriter.getF64FloatAttr(1));
+      rewriter.replaceOpWithNewOp<AtenSubTensorOp>(op, resultTy, self, trunc,
+                                                   alpha);
+      return success();
+    }
+
+    return failure();
+  }
+};
+} // namespace
+
 namespace {
 // decompose `fmod(x, y)` to `x - trunc(x/y) * y`
 class DecomposeAtenFmodTensorOp : public OpRewritePattern<AtenFmodTensorOp> {
@@ -10262,6 +10290,7 @@ class DecomposeComplexOpsPass
     addPatternIfTargetOpIsIllegal<DecomposeAtenRad2degOp>(patterns);
     addPatternIfTargetOpIsIllegal<DecomposeAtenCosineSimilarityOp>(patterns);
     addPatternIfTargetOpIsIllegal<DecomposeAtenTruncOp>(patterns);
+    addPatternIfTargetOpIsIllegal<DecomposeAtenFracOp>(patterns);
     addPatternIfTargetOpIsIllegal<DecomposeAtenFmodTensorOp>(patterns);
     addPatternIfTargetOpIsIllegal<DecomposeAtenBaddbmmOp>(patterns);
     addPatternIfTargetOpIsIllegal<DecomposeAtenFloorDivideOp>(patterns);

lib/Dialect/Torch/Transforms/LowerToBackendContract.cpp

@@ -541,6 +541,7 @@ static void markDecomposedOpsAsIllegal(MLIRContext *context,
   target.addIllegalOp<AtenRad2degOp>();
   target.addIllegalOp<AtenCosineSimilarityOp>();
   target.addIllegalOp<AtenTruncOp>();
+  target.addIllegalOp<AtenFracOp>();
   target.addIllegalOp<AtenNewEmptyStridedOp>();
   target.addIllegalOp<AtenEmptyStridedOp>();
   target.addIllegalOp<AtenBucketizeTensorOp>();

projects/pt1/e2e_testing/xfail_sets.py

@@ -3245,6 +3245,8 @@
     "Rot90MultipleRotationsModule_basic",
     "Rot90NegativeEvenRotationsModule_basic",
     "Rot90NegativeOddRotationsModule_basic",
+    # Error: 'aten::frac' to ONNX opset version 20 is not supported
+    "ElementwiseFracModule_basic",
     # Failure - unknown
     "BernoulliModule_basic",
     "Conv_Transpose1dModule_basic",

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

@@ -313,6 +313,9 @@ def aten〇ceil〡shape(self: List[int]) -> List[int]:
 def aten〇trunc〡shape(self: List[int]) -> List[int]:
     return upstream_shape_functions.unary(self)
 
+def aten〇frac〡shape(self: List[int]) -> List[int]:
+    return upstream_shape_functions.unary(self)
+
 def aten〇log〡shape(self: List[int]) -> List[int]:
     return upstream_shape_functions.unary(self)
 
@@ -2897,6 +2900,9 @@ def aten〇trunc〡dtype(self_rank_dtype: Tuple[int, int]) -> int:
     self_rank, self_dtype = self_rank_dtype
     return self_dtype
 
+def aten〇frac〡dtype(self_rank_dtype: Tuple[int, int]) -> int:
+    return torch.float32
+
 @check_dtype_function(_check_tensors_with_the_same_dtype(num_of_tensors=1, max=0))
 def aten〇clamp_max〡dtype(self_rank_dtype: Tuple[int, int], max: Union[int, float, complex]) -> int:
     self_rank, self_dtype = self_rank_dtype

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

@@ -507,6 +507,7 @@ def emit_with_mutating_variants(key, **kwargs):
     emit("aten::hardshrink : (Tensor, Scalar) -> (Tensor)")
     emit("aten::softshrink : (Tensor, Scalar) -> (Tensor)")
     emit("aten::polar : (Tensor, Tensor) -> (Tensor)")
+    emit("aten::frac : (Tensor) -> (Tensor)")
 
     # Ops with dynamic number of outputs
     emit("aten::unbind_copy.int : (Tensor, int) -> (Tensor[])")

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

@@ -2839,6 +2839,29 @@ def ElementwiseTruncIntModule_basic(module, tu: TestUtils):
 # ==============================================================================
 
 
+class ElementwiseFracModule(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    @export
+    @annotate_args(
+        [
+            None,
+            ([1, 6], torch.float32, True),
+        ]
+    )
+    def forward(self, a):
+        return torch.frac(a)
+
+
+@register_test_case(module_factory=lambda: ElementwiseFracModule())
+def ElementwiseFracModule_basic(module, tu: TestUtils):
+    module.forward(torch.tensor([[-torch.inf, torch.inf, torch.nan, -2.3, 0.0, 1.5]]))
+
+
+# ==============================================================================
+
+
 class ElementwiseSignModule(torch.nn.Module):
     def __init__(self):
         super().__init__()