"transforms/post_quantize.cc",
        "transforms/prepare_quantize.cc",
        "transforms/prepare_quantize_dynamic_range.cc",
        "transforms/prepare_quantize_helper.cc",
        "transforms/quantize.cc",
        "transforms/quantize_variables.cc",
        "utils/generated_op_quant_spec_getters.inc",
    ],
    hdrs = [
        "transforms/passes.h",
        "transforms/prepare_quantize_helper.h",
    ],

  UNIQUE = 103,
  CEIL = 104,
  REVERSE_V2 = 105,
  ADD_N = 106,
  GATHER_ND = 107,
  COS = 108,
  WHERE = 109,
  RANK = 110,
  ELU = 111,
  REVERSE_SEQUENCE = 112,
  MATRIX_DIAG = 113,
  QUANTIZE = 114,
  MATRIX_SET_DIAG = 115,
  ROUND = 116,
  HARD_SWISH = 117,
  IF = 118,
  WHILE = 119,
  NON_MAX_SUPPRESSION_V4 = 120,
  NON_MAX_SUPPRESSION_V5 = 121,
  SCATTER_ND = 122,
  SELECT_V2 = 123,

// RUN: stablehlo-quant-opt %s -split-input-file -stablehlo-quantize -verify-each=false | FileCheck %s

module attributes {tf_saved_model.semantics} {
  // CHECK-LABEL: same_scale_after_composite
  // CHECK-SAME: %[[ARG0:.*]]: tensor<1x2xf32>
  // CHECK-SAME: %[[ARG1:.*]]: tensor<2x3xf32>
  func.func private @same_scale_after_composite(%arg0: tensor<1x2xf32>, %arg1: tensor<2x3xf32>) -> tensor<3x1xf32> {

  %cst = arith.constant dense<1.5> : tensor<3xf32>
  %cst_0 = arith.constant dense<[1.0, 2.0, 3.0]> : tensor<3xf32>
  %w = arith.constant dense<2.0> : tensor<3x3x3x3xf32>
  %q = "tfl.quantize"(%w) {qtype = tensor<3x3x3x3x!quant.uniform<i8<-127:127>:f32:0,{1.0,2.0,3.0}>>} : (tensor<3x3x3x3xf32>) -> tensor<3x3x3x3x!quant.uniform<i8<-127:127>:f32:0,{1.0,2.0,3.0}>>

  // before converting TF_Conv to TFL_Conv
  (void)applyPatternsAndFoldGreedily(func, std::move(patterns));

  // Remove the wrapper of the tf.FakeQuant* ops and also insert the
  // tfl.quantize and tfl.dequantize to preserve the quantization parameters.
  // This is done after the first round of optimization to make sure all the
  // min/max operands of the tf.FakeQuant* are constants to be matched. The

      $_state.addAttribute("compressed_data", compressed_data);
    }]>
  ];
}

def TFL_QuantizeOp: TFL_Op<"quantize", [
    FirstAttrDerivedResultType,
    SameOperandsAndResultShape, NoMemoryEffect]> {
  let summary = "Quantize operator";

  let description = [{
    Converts floating point tensors to quantized integer tensors according to

    }
  }];
}

def TF_FakeQuantWithMinMaxVarsOp : TF_Op<"FakeQuantWithMinMaxVars", [Pure]> {
  let summary = [{
Fake-quantize the 'inputs' tensor of type float via global float scalars
  }];

  let description = [{
Fake-quantize the `inputs` tensor of type float via global float scalars
`min` and `max` to `outputs` tensor of same shape as `inputs`.

Attributes

// CHECK-LABEL: testQuantize
func.func @testQuantize(tensor<? x f32>) -> tensor<? x !quant.uniform<u8:f32, 0.1:128>> {
^bb0(%arg0: tensor<? x f32>):
  // CHECK: %0 = "tfl.quantize"(%arg0) <{qtype = tensor<?x!quant.uniform<u8:f32, 1.000000e-01:128>>}>
  %0 = "tfl.quantize"(%arg0) {qtype = tensor<? x !quant.uniform<u8:f32, 0.1:128>>} : (tensor<? x f32>) -> tensor<? x !quant.uniform<u8:f32, 0.1:128>>
  func.return %0 : tensor<? x !quant.uniform<u8:f32, 0.1:128>>

// quotient is the exclusive OR of the operands’ signs; the sign of a sum,
// or of a difference x−y regarded as a sum x+(−y), differs from at most
// one of the addends’ signs; and the sign of the result of conversions,
// the quantize operation, the roundToIntegral operations, and the
// roundToIntegralExact (see 5.3.1) is the sign of the first or only operand.
// These rules shall apply even when operands or results are zero or infinite.
//

    "UNIQUE",
    "CEIL",
    "REVERSE_V2",
    "ADD_N",
    "GATHER_ND",
    "COS",
    "WHERE",
    "RANK",
    "ELU",
    "REVERSE_SEQUENCE",
    "MATRIX_DIAG",
    "QUANTIZE",
    "MATRIX_SET_DIAG",
    "ROUND",
    "HARD_SWISH",
    "IF",
    "WHILE",
    "NON_MAX_SUPPRESSION_V4",
    "NON_MAX_SUPPRESSION_V5",
    "SCATTER_ND",
    "SELECT_V2",