}

  return false;
}

// Returns true if the value's element type is F32.
bool IsF32Value(Value value) {
  return mlir::cast<ShapedType>(value.getType()).getElementType().isF32();
}

// Returns the number of elements in attr if it is a static shape, 1 otherwise,
// as an unranked int32 Attribute.
TypedAttr GetNumElementsOrOne(Type type) {
  auto shaped_type = mlir::cast<ShapedType>(type);

             .isF32()) {
      return failure();
    }
    MLIRContext *context = rewriter.getContext();
    llvm::SmallVector<Value, 2> operands{op.getA(), op.getB()};
    for (Value &operand : operands) {
      TensorType tensor_type = mlir::cast<TensorType>(operand.getType());
      Type element_type = tensor_type.getElementType();
      if (element_type.isF32()) continue;

  return base_function_name.concat("_fn").str();
}

bool ContainsFloatResultType(ArrayRef<Type> result_types) {
  for (auto current_type : result_types) {
    if (mlir::dyn_cast<TensorType>(current_type).getElementType().isF32())
      return true;
  }
  return false;
}

// Unwraps quantization parameters of PartitionedCall ops with quantized
// input/outputs that are created from QuantizePass.
class QuantizeFunctionPattern

      if (!epsilon)
        epsilon = rewriter.getFloatAttr(rewriter.getF32Type(), 0.0001f);

      if (!(((mlir::isa<::mlir::FloatAttr>(epsilon))) &&
            ((mlir::cast<::mlir::FloatAttr>(epsilon).getType().isF32())))) {
        return rewriter.notifyMatchFailure(
            fused_batch_norm_op, [&](::mlir::Diagnostic &diag) {
              diag << "op 'tf.FusedBatchNormV3' attribute 'epsilon' failed to "

  LogicalResult matchAndRewrite(
      OpT op, typename OpT::Adaptor adaptor,
      ConversionPatternRewriter &rewriter) const override {
    Type dtype = op.getElementDtype();
    if (!(dtype.isF16() || dtype.isF32() || dtype.isF64() ||
          dtype.isInteger(1) || dtype.isInteger(8) || dtype.isInteger(16) ||
          dtype.isInteger(32) || dtype.isInteger(64))) {
      const char *error_info =

      return tflite::TensorType_UINT16;
    } else {
      return Status(absl::StatusCode::kInvalidArgument,
                    "'isSigned' can only be set for 8/16-bits integer type");
    }
  }

  if (type.isF32()) {
    return tflite::TensorType_FLOAT32;
  } else if (type.isF16()) {
    return tflite::TensorType_FLOAT16;
  } else if (type.isBF16()) {
    return tflite::TensorType_BFLOAT16;
  } else if (type.isF64()) {

  // For f32/f16 data type decision depends on the filter size in spatial
  // dimensions, for other data types we keep current data format.
  if (!input_ty.getElementType().isF32() && !input_ty.getElementType().isF16())
    return getDataFormat();

  // Keep current data format if filter rank is unknown or not equal to 4.
  auto filter_ty = mlir::dyn_cast<RankedTensorType>(getFilter().getType());

        return op.emitOpError()
               << "Missing " << attr_name << " attribute in backend_config";
      }
      auto attr = backend_config.getAs<FloatAttr>(attr_name);
      if (!attr || !attr.getType().isF32()) {
        return op.emitOpError()
               << attr_name
               << " attribute in backend_config must be of f32 type";
      }
      return success();
    };
    auto check_bool_attr =

   (TF_ConstOp:$one (GetScalarOfType<1> $num_lower)),
   (TF_ConstOp:$neg_one (GetScalarOfType<-1> $num_lower)),
   (GetDimensionSize<-2>:$m $input, (IsI32 $num_lower)),
   (GetDimensionSize<-1>:$n $input, (IsI32 $num_upper)),
   (TF_SelectV2Op:$num_lower_or_m (TF_LessOp $num_lower, $zero),
                                  $m, $num_lower),
   (TF_SelectV2Op:$num_upper_or_n (TF_LessOp $num_upper, $zero),

			return range_.Stride == 1 || (r-range_.Lo)%range_.Stride == 0
		}
		if r < range_.Lo {
			hi = m
		} else {
			lo = m + 1
		}
	}
	return false
}

// is32 reports whether r is in the sorted slice of 32-bit ranges.
func is32(ranges []Range32, r uint32) bool {
	if len(ranges) <= linearMax {
		for i := range ranges {
			range_ := &ranges[i]
			if r < range_.Lo {
				return false
			}