%cst_1 = "tf.Const"() {value = dense<3.000000e+00> : tensor<2x3x512x2xf32>} : () -> tensor<2x3x512x2xf32>
    %1 = "tf.PartitionedCall"(%arg0, %cst_1) {_tfl_quant_trait = "fully_quantizable", config = "", config_proto = "", executor_type = "", f = @composite_conv2d_fn_1} : (tensor<1x3x4x512xf32>, tensor<2x3x512x2xf32>) -> tensor<*xf32>

  %cst = "tf.Const"() {value = dense<3.000000e+00> : tensor<2x3x512x512xf32>} : () -> tensor<2x3x512x512xf32>
  %0 = "tf.Conv2D"(%arg0, %cst) {
    data_format = "NHWC", device = "", dilations = [1, 1, 1, 1], explicit_paddings = [],
    padding = "SAME", strides = [1, 1, 2, 1], use_cudnn_on_gpu = true
  } : (tensor<1x3x4x512xf32>, tensor<2x3x512x512xf32>) -> tensor<*xf32>
  func.return %0 : tensor<*xf32>

    %0 = "tf.FakeQuantWithMinMaxArgs"(%arg0) {device = "", max = 2.000000e-01 : f32, min = -1.000000e-01 : f32, narrow_range = false, num_bits = 8 : i64} : (tensor<1x3x4x3xf32>) -> tensor<1x3x4x3xf32>

    %0 = "quantfork.qcast"(%arg0) : (tensor<1x3x4x3xf32>) -> tensor<1x3x4x3x!quant.uniform<i8:f32, 0.0011764706057660721:-43>>

    %2 = "quantfork.stats"(%1) {layerStats = dense<[4.00000000e-6, 9.80000000e-1]> : tensor<2xf32>} : (tensor<2x3x2x2xf32>) -> tensor<2x3x2x2xf32>
    return %2 : tensor<2x3x2x2xf32>
  }
// Checks that the quantized XlaCallModule has been replaced by a CallOp, which
// calls the quantized entry function.

            index_vector_dim = 2>,
          slice_sizes = array<i64: 1, 2, 2>,
          indices_are_sorted = false
        } : (tensor<3x4x2xf32>, tensor<2x3x2xi64>) -> tensor<2x3x2x2xf32>
        func.return %2 : tensor<2x3x2x2xf32>
      }
    }
  )mlir";
  OwningOpRef<ModuleOp> module_op = ParseModuleOpString(kModuleGather);
  ASSERT_TRUE(module_op);

  func::FuncOp main_fn = FindMainFuncOp(*module_op);

    indices_are_sorted = false
  } : (tensor<3x4x2xf32>, tensor<2x3x2xi64>) -> tensor<2x3x2x2xf32>
    %8 = "quantfork.qcast"(%7) {volatile} : (tensor<2x3x2x2xf32>) -> tensor<2x3x2x2x!quant.uniform<i8:f32, 0.13170163023705575:-1>>
    %9 = "quantfork.dcast"(%8) : (tensor<2x3x2x2x!quant.uniform<i8:f32, 0.13170163023705575:-1>>) -> tensor<2x3x2x2xf32>
    return %9 : tensor<2x3x2x2xf32>
  }

  // CHECK: quantized_dot_general_fn_1

    disabled_checks = [], has_token_input_output = false, module = "",
    platforms = [], version = 5 : i64
  } : (tensor<3x4x2xf32>, tensor<2x3x2xi64>) -> tensor<2x3x2x2xf32>
  return %0 : tensor<2x3x2x2xf32>
}

// CHECK-LABEL: func.func @no_qdq_except_conv_and_dot_general
// CHECK-NOT: quantfork.qcast
// CHECK-NOT: quantfork.dcast

// -----

    slice_sizes = array<i64: 1, 1, 2, 2>,
    indices_are_sorted = false
  } : (tensor<3x4x2x2xf32>, tensor<2x3x2xi64>) -> tensor<2x3x2x2xf32>
  return %0 : tensor<2x3x2x2xf32>
}

// CHECK-LABEL: func @gather_float
// CHECK: stablehlo.gather
// CHECK-NOT: tfl.gather_nd
// CHECK-NOT: tfl.gather

// -----

// Tests that a quantized `stablehlo.dynamic_slice` is converted to `tfl.slice`.