input_to_cell_intermediate = tensor<!quant.calibrated<f32<-4.000000e+00:4.000000e+00>>>,
      input_to_forget_intermediate = tensor<!quant.calibrated<f32<-1.600000e+01:1.600000e+01>>>,
      input_to_output_intermediate = tensor<!quant.calibrated<f32<-1.000000e+00:1.000000e+00>>>,
      proj_clip = 0.000000e+00 : f32,time_major = false} : (
        tensor<1x5xf32>,
        none, tensor<2x5xf32>, tensor<2x5xf32>, tensor<2x5xf32>,

= tensor<0x!quant.uniform<i16:f32, 0.0049890000373125076>>, input_to_forget_intermediate = tensor<0x!quant.uniform<i16:f32, 0.0078849997371435165>>, input_to_cell_intermediate = tensor<0x!quant.uniform<i16:f32, 0.0087630003690719604>>, input_to_output_intermediate = tensor<0x!quant.uniform<i16:f32, 0.0057529998011887074>>, effective_hidden_scale_intermediate = tensor<0x!quant.uniform<i8:f32, 0.0075630000792443752:2>>, kernel_type = #tfl<lstm_kernel_type_attr FULL>, proj_clip = 0.01 : f32} : (ten...

// CHECK-SAME: input_to_forget_intermediate = tensor<*x!quant.calibrated<f32<-1.600000e+01:1.600000e+01>>>
// CHECK-SAME: input_to_input_intermediate = tensor<*x!quant.calibrated<f32<-3.200000e+01:3.200000e+01>>>
// CHECK-SAME: input_to_output_intermediate = tensor<*x!quant.calibrated<f32<-1.000000e+00:1.000000e+00>>>

// Checks if calibrated type is exported back to quantized type.
// RoundTrip: name: "effective_hidden_scale_intermediate",

      fused_activation_function = "TANH",
      input_to_cell_intermediate = tensor<0xf32>,
      input_to_forget_intermediate = tensor<0xf32>,
      input_to_input_intermediate = tensor<0xf32>,
      input_to_output_intermediate = tensor<0xf32>,
      proj_clip = 0.000000e+00 : f32,
      time_major = false} : (
        tensor<1x2x3xf32>,
        tensor<1x1xf32>, tensor<1x1xf32>, tensor<1x1xf32>, tensor<1x1xf32>,

// CHECK-NEXT:       name: "input_to_cell_intermediate",
// CHECK-NEXT:       has_rank: true
// CHECK-NEXT:     }, {
// CHECK-NEXT:       shape: [ 0 ],
// CHECK-NEXT:       name: "input_to_output_intermediate",
// CHECK-NEXT:       has_rank: true
// CHECK-NEXT:     }, {
// CHECK-NEXT:       shape: [ 0 ],
// CHECK-NEXT:       type: INT8,
// CHECK-NEXT:       name: "effective_hidden_scale_intermediate",

= tensor<0x!quant.uniform<i16:f32, 0.0049890000373125076>>, input_to_forget_intermediate = tensor<0x!quant.uniform<i16:f32, 0.0078849997371435165>>, input_to_cell_intermediate = tensor<0x!quant.uniform<i16:f32, 0.0087630003690719604>>, input_to_output_intermediate = tensor<0x!quant.uniform<i16:f32, 0.0057529998011887074>>, effective_hidden_scale_intermediate = tensor<0x!quant.uniform<i8<-127:127>:f32, 0.0075630000792443752:2>>, kernel_type = #tfl<lstm_kernel_type_attr FULL>, proj_clip = 0.01 : f32}...

      /*input_to_input_intermediate=*/mlir::TypeAttr(),
      /*input_to_forget_intermediate=*/mlir::TypeAttr(),
      /*input_to_cell_intermediate=*/mlir::TypeAttr(),
      /*input_to_output_intermediate=*/mlir::TypeAttr(),
      /*effective_hidden_scale_intermediate=*/mlir::TypeAttr());

  // Cast the static shaped lstm result to FuncOp's signature -
  // Ranked but unknown 2nd dimension to support stacking these.

constexpr const char* intermediate_attributes[] = {
    "input_to_input_intermediate", "input_to_forget_intermediate",
    "input_to_cell_intermediate", "input_to_output_intermediate",
    "effective_hidden_scale_intermediate"};

// Calculates the minimum power of two that is not less than the value.
double PowerOfTwoBound(double value);

tensor<0x!quant.uniform<i8<-127:127>:f32, 0.0077881771139800549>>, fused_activation_function = "TANH", input_to_cell_intermediate = tensor<0xf32>, input_to_forget_intermediate = tensor<0xf32>, input_to_input_intermediate = tensor<0xf32>, input_to_output_intermediate = tensor<0xf32>, proj_clip = 0.000000e+00 : f32, time_major = false}> : (tensor<?x?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?x?xf32>, tensor<?xf32>,...

    // Create intermediate value

    const llvm::SmallVector<llvm::StringRef, 5> kIntermediateNames = {
        "input_to_input_intermediate", "input_to_forget_intermediate",
        "input_to_cell_intermediate", "input_to_output_intermediate",
        "effective_hidden_scale_intermediate"};
    for (auto type_and_name :
         llvm::zip(intermediate_types, kIntermediateNames)) {
      mlir::TypeAttr type_attr =