[onnx] Support `fp8` for `onnx.QuantizeLinear` (#3619)

We need to directly decompose quantize linear for `fp8` types as the
equivalent torch operations do not support the operation.

lib/Conversion/TorchOnnxToTorch/DefaultDomainQtoZ.cpp

@@ -214,6 +214,7 @@ void mlir::torch::onnx_c::populateDefaultDomainQtoZ(
             binder.tensorResultType(resultType))
           return failure();
 
+        auto loc = binder.getLoc();
         Value operand = operands[0];
         Value scale = operands[1];
         Value zeropoint = operands[2];
@@ -225,7 +226,16 @@ void mlir::torch::onnx_c::populateDefaultDomainQtoZ(
           return rewriter.notifyMatchFailure(binder.op,
                                              "requires known result dtype");
 
-        if (scaleTy.getSizes().size() == 0) {
+        auto resultETy = resultType.getDtype();
+
+        bool rank0 = scaleTy.getSizes().size() == 0;
+        bool length1 =
+            scaleTy.getSizes().size() == 1 && scaleTy.getSizes()[0] == 1;
+
+        if (!rank0 && !length1)
+          return rewriter.notifyMatchFailure(binder.op,
+                                             "unimplemented: non-scalar scale");
+
         auto qTensorTy = getQTorchTypeFromTorchIntType(resultType);
         if (!qTensorTy) {
           return rewriter.notifyMatchFailure(binder.op,
@@ -235,23 +245,42 @@ void mlir::torch::onnx_c::populateDefaultDomainQtoZ(
         auto torchqTy = Torch::getScalarTypeForType(qTensorTy.getDtype());
 
         Value tyConst = rewriter.create<Torch::ConstantIntOp>(
-              binder.getLoc(), rewriter.getType<Torch::IntType>(),
+            loc, rewriter.getType<Torch::IntType>(),
             rewriter.getIntegerAttr(rewriter.getIntegerType(64),
                                     static_cast<int64_t>(torchqTy)));
 
         scale = rewriter.create<Torch::AtenItemOp>(
-              binder.getLoc(), rewriter.getType<Torch::FloatType>(), scale);
+            loc, rewriter.getType<Torch::FloatType>(), scale);
-          zeropoint = rewriter.create<Torch::AtenItemOp>(
-              binder.getLoc(), rewriter.getType<Torch::IntType>(), zeropoint);
 
-          auto quantize = rewriter.create<Torch::AtenQuantizePerTensorOp>(
+        bool fpResult = isa<mlir::FloatType>(resultETy);
-              binder.getLoc(), qTensorTy, operand, scale, zeropoint, tyConst);
+        Type zeropointTy = rewriter.getType<Torch::IntType>();
-          rewriter.replaceOpWithNewOp<Torch::AtenIntReprOp>(
+        if (fpResult)
-              binder.op, resultType, quantize);
+          zeropointTy = rewriter.getType<Torch::FloatType>();
+        zeropoint =
+            rewriter.create<Torch::AtenItemOp>(loc, zeropointTy, zeropoint);
+
+        if (fpResult) {
+          Value none = rewriter.create<Torch::ConstantNoneOp>(loc);
+          Value cstFalse = rewriter.create<Torch::ConstantBoolOp>(loc, false);
+          Value one = rewriter.create<Torch::ConstantFloatOp>(
+              loc, rewriter.getF64FloatAttr(1.0));
+          Value div = rewriter.create<Torch::AtenDivScalarOp>(
+              loc, operand.getType(), operand, scale);
+          Value add = rewriter.create<Torch::AtenAddScalarOp>(
+              loc, operand.getType(), div, zeropoint, one);
+
+          rewriter.replaceOpWithNewOp<Torch::AtenToDtypeOp>(
+              binder.op, resultType, add, tyConst,
+              /*non_blocking=*/cstFalse, /*copy=*/cstFalse,
+              /*memory_format=*/none);
           return success();
         }
 
-        return failure();
+        auto quantize = rewriter.create<Torch::AtenQuantizePerTensorOp>(
+            loc, qTensorTy, operand, scale, zeropoint, tyConst);
+        rewriter.replaceOpWithNewOp<Torch::AtenIntReprOp>(binder.op, resultType,
+                                                          quantize);
+        return success();
       });
   patterns.onOp(
       "QLinearConv", 1,

test/Conversion/TorchOnnxToTorch/simple_ops_q_to_z.mlir

@@ -47,6 +47,23 @@ func.func @test_quantizelinear_i32(%arg0: !torch.vtensor<[6],f32>, %arg1: !torch
 
 // -----
 
+// CHECK-LABEL: @test_quantizelinear_f8
+func.func @test_quantizelinear_f8(%arg0: !torch.vtensor<[6],f32>, %arg1: !torch.vtensor<[],f32>, %arg2: !torch.vtensor<[],f32>) -> !torch.vtensor<[6],f8E4M3FN> attributes {torch.onnx_meta.ir_version = 9 : si64, torch.onnx_meta.opset_version = 19 : si64} {
+  // CHECK: %[[DTYPE:.+]] = torch.constant.int 24
+  // CHECK: %[[SCALE:.+]] = torch.aten.item %arg1 : !torch.vtensor<[],f32> -> !torch.float
+  // CHECK: %[[ZP:.+]] = torch.aten.item %arg2 : !torch.vtensor<[],f32> -> !torch.float
+  // CHECK: %[[NONE:.+]] = torch.constant.none
+  // CHECK: %[[FALSE:.+]] = torch.constant.bool false
+  // CHECK: %[[ONE:.+]] = torch.constant.float 1.000000e+00
+  // CHECK: %[[DIV:.+]] = torch.aten.div.Scalar %arg0, %[[SCALE]]
+  // CHECK: %[[ADD:.+]] = torch.aten.add.Scalar %[[DIV]], %[[ZP]], %[[ONE]]
+  // CHECK: %[[CAST:.+]] = torch.aten.to.dtype %[[ADD]], %[[DTYPE]], %[[FALSE]], %[[FALSE]], %[[NONE]]
+  %0 = torch.operator "onnx.QuantizeLinear"(%arg0, %arg1, %arg2) : (!torch.vtensor<[6],f32>, !torch.vtensor<[],f32>, !torch.vtensor<[],f32>) -> !torch.vtensor<[6],f8E4M3FN>
+  return %0 : !torch.vtensor<[6],f8E4M3FN>
+}
+
+// -----
+
 // CHECK-LABEL: @test_qlinearconv_nobias
 func.func @test_qlinearconv_nobias(%arg0: !torch.vtensor<[1,1,7,7],ui8>, %arg1: !torch.vtensor<[],f32>, %arg2: !torch.vtensor<[],ui8>, %arg3: !torch.vtensor<[1,1,1,1],ui8>, %arg4: !torch.vtensor<[1],f32>, %arg5: !torch.vtensor<[1],ui8>, %arg6: !torch.vtensor<[],f32>, %arg7: !torch.vtensor<[],ui8>) -> !torch.vtensor<[1,1,7,7],ui8> attributes {torch.onnx_meta.ir_version = 5 : si64, torch.onnx_meta.opset_version = 10 : si64, torch.onnx_meta.producer_name = "backend-test", torch.onnx_meta.producer_version = ""} {
   %0 = torch.operator "onnx.QLinearConv"(%arg0, %arg1, %arg2, %arg3, %arg4, %arg5, %arg6, %arg7) : (!torch.vtensor<[1,1,7,7],ui8>, !torch.vtensor<[],f32>, !torch.vtensor<[],ui8>, !torch.vtensor<[1,1,1,1],ui8>, !torch.vtensor<[1],f32>, !torch.vtensor<[1],ui8>, !torch.vtensor<[],f32>, !torch.vtensor<[],ui8>) -> !torch.vtensor<[1,1,7,7],ui8>