// CHECK-DAG: %[[input_14:.*]] = "tfl.dequantize"({{.*}}) : (tensor<2x!quant.uniform<i32:f32, 1.1827185827747504E-4>>) -> tensor<2xf32>
// CHECK-DAG: %[[input_15:.*]] = "tfl.dequantize"({{.*}}) : (tensor<2x!quant.uniform<i32:f32, 1.8229184289320815E-4>>) -> tensor<2xf32>
// CHECK-DAG: %[[input_16:.*]] = "tfl.dequantize"({{.*}}) : (tensor<4x2x!quant.uniform<i8<-127:127>:f32, 0.013545581674951268>>) -> tensor<4x2xf32>

  %0 = "tfl.pseudo_qconst"() { qtype = tensor<3x!quant.uniform<i8:f32, 0.1>>, value = dense<2> : tensor<3xi8>} : () -> tensor<3x!quant.uniform<i8:f32, 0.1>>
  %1 = "tfl.mul"(%arg0, %0) {fused_activation_function = "NONE"} : (tensor<3x!quant.uniform<i8:f32, 0.1>>, tensor<3x!quant.uniform<i8:f32, 0.1>>) -> tensor<3x!quant.uniform<i8:f32, 0.1>> loc("mul")
  func.return %1 : tensor<3x!quant.uniform<i8:f32, 0.1>>

  %0 = "tfl.pseudo_qconst"() { qtype = tensor<3x!quant.uniform<i8:f32, 1.0>>, value = dense<2> : tensor<3xi8>} : () -> tensor<3x!quant.uniform<i8:f32, 1.0>>
  %1 = "tfl.mul"(%arg0, %0) {fused_activation_function = "NONE"} : (tensor<3x!quant.uniform<i8:f32, 1.0>>, tensor<3x!quant.uniform<i8:f32, 1.0>>) -> tensor<3x!quant.uniform<i8:f32, 1.0>> loc("mul")
  func.return %1 : tensor<3x!quant.uniform<i8:f32, 1.0>>

  %0 = "tfl.pseudo_qconst"() { qtype = tensor<3x3x!quant.uniform<i32:f32:0, {1.0, 0.5:1, 0.25:1}>>, value = dense<1> : tensor<3x3xi32>} : () -> tensor<3x3x!quant.uniform<i32:f32:0, {1.0, 0.5:1, 0.25:1}>>
  func.return %0 : tensor<3x3x!quant.uniform<i32:f32:0, {1.0, 0.5:1, 0.25:1}>>
}

func.func @qu8() -> tensor<3x!quant.uniform<u8<1:255>:f32, 1.0>> {
  // CHECK-LABEL: @qu8

    %1 = "quantfork.dcast"(%0) : (tensor<2x2x3x2x!quant.uniform<i8<-127:127>:f32:3, {4.000000e-03,5.000000e-03}>>) -> tensor<*xf32>
    %2 = "quantfork.qcast"(%arg0) : (tensor<1x2x2x3xf32>) -> tensor<1x2x2x3x!quant.uniform<i8:f32, 8.000000e-03>>
    %3 = "quantfork.dcast"(%2) : (tensor<1x2x2x3x!quant.uniform<i8:f32, 8.000000e-03>>) -> tensor<*xf32>

  %5 = "tfl.reshape"(%4, %0) : (tensor<1x112x112x32x!quant.uniform<u8:f32, 0.023528476789885875>>, tensor<2xi32>) -> tensor<1x401408x!quant.uniform<u8:f32, 0.023528476789885875>>
  %6 = "tfl.softmax"(%5) {beta = 1.000000e+00 : f32} : (tensor<1x401408x!quant.uniform<u8:f32, 0.023528476789885875>>) -> tensor<1x401408x!quant.uniform<u8:f32, 3.906250e-03>>

// CHECK: %[[CST:.+]] = stablehlo.constant() <{value = dense<127> : tensor<2x3xi8>}> : () -> tensor<2x3x!quant.uniform<i8<-127:127>:f32, 0.0023622048182750312>>
// CHECK: %[[CALL:.+]] = call @quantized_dot_general_fn(%[[ARG0]], %[[CST]]) {_quantization_method = "weight_only_ptq { }"} : (tensor<1x2xf32>, tensor<2x3x!quant.uniform<i8<-127:127>:f32, 0.0023622048182750312>>) -> tensor<1x3xf32>
// CHECK: return %[[CALL]]

// CHECK: quantized_dot_general_fn

  // CHECK: %[[cst:.*]] = stablehlo.constant
  // CHECK: %[[q1:.*]] = "quantfork.qcast"(%[[cst]])
  // CHECK-SAME: quant.uniform<i8:f32, 0.0040316890267764818:127>
  // CHECK: %[[dq1:.*]] = "quantfork.dcast"(%[[q1]])
  // CHECK-SAME: quant.uniform<i8:f32, 0.0040316890267764818:127>
  %cst = stablehlo.constant dense<[[-0.960978984, -0.390246302], [-0.790828585, -0.601039409], [-1.0280807, -1.02731466]]> : tensor<3x2xf32>

  // CHECK-DAG: %[[LHS:.*]] = mhlo.uniform_quantize %arg0 : (tensor<3x2xf32>) -> tensor<3x2x!quant.uniform<i32:f32, 2.000000e+00:4>>
  // CHECK-DAG: %[[LHS1:.*]] = mhlo.bitcast_convert %[[LHS]] : (tensor<3x2x!quant.uniform<i32:f32, 2.000000e+00:4>>) -> tensor<3x2xi32>
  // CHECK-DAG: %[[LHS2:.*]] = mhlo.bitcast_convert %[[LHS1]] : (tensor<3x2xi32>) -> tensor<3x2x!quant.uniform<i32:f32, 2.000000e+00:4>>
  %0 = "tf.UniformQuantize"(%input, %input_scales, %input_zps) {

// CustomOp-DAG: %[[q_b:.*]] = "tfl.quantize"(%[[b]]) <{qtype = tensor<2048x1x1x1x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>}> : (tensor<2048x1x1x1xf32>) -> tensor<2048x1x1x1x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>