func.return %4 : tensor<*xf32>
}

// CHECK: %[[q:.*]] = "quantfork.qcast"(%arg0)
// CHECK: %[[sc1:.*]] = "quantfork.scast"(%[[q]]) : (tensor<*x!quant.uniform<i8:f32, 5.000000e-02:-10>>)
// CHECK: %[[fcast:.*]] = "tf.Cast"(%[[sc1]]) <{Truncate = false}> : (tensor<*xi8>) -> tensor<*xf32>
// CHECK: %[[avgpool_f32:.*]] = "tf.AvgPool"(%[[fcast]])
// CHECK-SAME: (tensor<*xf32>) -> tensor<*xf32>
// CHECK: %[[round:.*]] = "tf.Round"(%[[avgpool_f32]])

// CHECK: %[[scast:.*]] = "quantfork.scast"(%[[conv]]
// CHECK: %[[fcast:.*]] = "tf.Cast"(%[[scast]]) <{Truncate = false}> : (tensor<*xi8>) -> tensor<*xf32>
// CHECK: %[[avgpool_f32:.*]] = "tf.AvgPool"(%[[fcast]])
// CHECK-SAME: (tensor<*xf32>) -> tensor<*xf32>
// CHECK: %[[round:.*]] = "tf.Round"(%[[avgpool_f32]])
// CHECK: %[[icast:.*]] = "tf.Cast"(%[[round]]) <{Truncate = false}> : (tensor<*xf32>) -> tensor<*xi8>

     *
     * @param outputType The type to cast the input to
     * @param object The object to be cast (must not be {@code null})
     * @param <O> The type to be cast to
     * @param <I> The type of the object to be vast
     * @return The input object, cast to the output type
     */
    public static <O, I> O cast(Class<O> outputType, I object) {
        try {

// A QuantizeCast (qcast) represents a potential type shift from a quantizable
// type to a quantized type.
//
// At runtime, a qcast will apply the transformation expressed by its
// operand and result type. For flexibility during transformation, it is also
// possible to have a qcast that performs no transformation (both its
// operand and result type are quantizable).
//
// A qcast will typically originate from either:

  }
};

// Inserts a "quant.qcast" and "quant.dcast" op pair (QDQs) in place of the
// tf.FakeQyantWithMinMax{Vars|VarsPerChannel|Args}Op
// before the op being constant folded. Since the constant
// folding logic will use a "arith.constant" op to replace the
// "tf.FakeQuantWithMinMaxVarsOp", the "quant.qcast" op is used to preserve
// the quantization parameters as a TypeAttr and "quant.dcast" op used to

  // CHECK-NOT: %[[q0:.*]] = "quantfork.qcast"(%[[cst0]])
  // CHECK-NOT: %[[q1:.*]] = "quantfork.qcast"(%[[cst1]])
  // CHECK: %[[q2:.*]] = "quantfork.qcast"(%[[cst2]])
  // CHECK-SAME: quant.uniform<i8:f32, 0.023529411764705882:-128>
  // CHECK: %[[dq2:.*]] = "quantfork.dcast"(%[[q2]])
  // CHECK-SAME: quant.uniform<i8:f32, 0.023529411764705882:-128>
  // CHECK: %[[q3:.*]] = "quantfork.qcast"(%[[cst3]])

    %5 = "quantfork.dcast"(%4) : (tensor<2x3x1024x3x!quant.uniform<i8<-127:127>:f32, 4.000000e-01>>) -> tensor<2x3x1024x3xf32>
    %6 = "quantfork.qcast"(%arg0) {volatile} : (tensor<2x3x1x1024xf32>) -> tensor<2x3x1x1024x!quant.uniform<i8:f32, 5.000000e-01:2>>

    // CHECK: %[[DQ:.*]] = "quantfork.dcast"(%[[RESHAPE]]) : (tensor<3x1x!quant.uniform<i8:f32, 0.13170163023705575:-1>>) -> tensor<3x1xf32>
    // CHECK: return %[[DQ]]
    %0 = "quantfork.qcast"(%arg0) {volatile} : (tensor<1x2xf32>) -> tensor<1x2x!quant.uniform<i8:f32, 5.000000e-03>>
    %1 = "quantfork.dcast"(%0) : (tensor<1x2x!quant.uniform<i8:f32, 5.000000e-03>>) -> tensor<1x2xf32>

   %1 = "quantfork.qcast"(%0) {volatile} : (tensor<2x3xf32>) -> tensor<2x3x!quant.uniform<i8<-127:127>:f32, 5.000000e-03>>
   %2 = "quantfork.dcast"(%1) : (tensor<2x3x!quant.uniform<i8<-127:127>:f32, 5.000000e-03>>) -> tensor<2x3xf32>
   %3 = "quantfork.qcast"(%arg0) {volatile} : (tensor<1x2xf32>) -> tensor<1x2x!quant.uniform<i8:f32, 6.000000e-03:-128>>

  // CHECK-NOT: "quantfork.qcast"
  // CHECK-NOT: "quantfork.dcast"
  // CHECK: return %[[CST]]
  %cst = stablehlo.constant dense<[[-0.960978984, -0.390246302], [-0.790828585, -0.601039409], [-1.0280807, -1.02731466]]> : tensor<3x2xf32>
  %q = "quantfork.qcast"(%cst) {volatile} : (tensor<3x2xf32>) -> tensor<3x2x!quant.uniform<i8:f32, 0.013075299590241675:-64>>