TensorFlow Lite operator versions

This document describes TensorFlow Lite's op versioning schema. Op versioning enables developers to add new functionalities and parameters into existing ops. In addition, it guarantees the following:

  • Backward compatibility: New TensorFlow Lite implementation should handle an old model file.
  • Forward compatibility: Old TensorFlow Lite implementation should handle a new model file produced by new version of converter, as long as no new features are used.
  • Forward in-compatibility detection: If an old TensorFlow Lite implementation reads a new model that contains a new version of an op which isn't supported, it should report the error.

Example: Adding dilation into depthwise convolution

The remainder of this document explains op versioning in TFLite by showing how to add dilation parameters to the depthwise convolution operation.

Knowledge of dilation is not required to understand this document. Note that:

  • 2 new integer parameters will be added: dilation_width_factor and dilation_height_factor.
  • Old depthwise convolution kernels that don't support dilation are equivalent to setting the dilation factors to 1.

Change FlatBuffer schema

To add new parameters into an op, change the options table in lite/schema/schema.fbs.

For example, the options table of depthwise convolution looks like this:

table DepthwiseConv2DOptions {
  padding:Padding;
  stride_w:int;
  stride_h:int;
  depth_multiplier:int;
  fused_activation_function:ActivationFunctionType;
}

When adding new parameters:

  • Add comments indicating which parameters are supported by which version.
  • When the new implementation gets the default values for newly added parameters, it should work exactly the same as the old implementation.

The table will be like this after the new parameters are added:

table DepthwiseConv2DOptions {
  // Parameters for DepthwiseConv version 1 or above.
  padding:Padding;
  stride_w:int;
  stride_h:int;
  depth_multiplier:int;
  fused_activation_function:ActivationFunctionType;
  // Parameters for DepthwiseConv version 2 or above.
  dilation_w_factor:int = 1;
  dilation_h_factor:int = 1;
}

The file lite/schema/schema_generated.h should be re-generated for the new schema.

Change C structures and kernel implementation

In TensorFlow Lite, the kernel implementation is decoupled from FlatBuffer definition. The kernels read the parameter from C structures defined in lite/c/builtin_op_data.h.

The original depthwise convolution parameter is as follows:

typedef struct {
  TfLitePadding padding;
  int stride_width;
  int stride_height;
  int depth_multiplier;
  TfLiteFusedActivation activation;
} TfLiteDepthwiseConvParams;

As with the FlatBuffer schema, add comments indicating which parameters are supported starting from which version. The result is seen below:

typedef struct {
  // Parameters for DepthwiseConv version 1 or above.
  TfLitePadding padding;
  int stride_width;
  int stride_height;
  int depth_multiplier;
  TfLiteFusedActivation activation;
  // Parameters for DepthwiseConv version 2 or above.
  int dilation_width_factor;
  int dilation_height_factor;
} TfLiteDepthwiseConvParams;

Please also change the kernel implementation to read the newly added parameters from the C structures. The details are omitted here.

Change the FlatBuffer reading code

The logic to read FlatBuffer and produce C structure is in lite/core/api/flatbuffer_conversions.cc.

Update the file to handle the new parameters, as shown below:

TfLiteStatus ParseDepthwiseConv2D(const Operator* op,
                                  ErrorReporter* error_reporter,
                                  BuiltinDataAllocator* allocator,
                                  void** builtin_data) {
  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);

  SafeBuiltinDataAllocator safe_allocator(allocator);

  std::unique_ptr<TfLiteDepthwiseConvParams,
                  SafeBuiltinDataAllocator::BuiltinDataDeleter>
      params = safe_allocator.Allocate<TfLiteDepthwiseConvParams>();
  TF_LITE_ENSURE(error_reporter, params != nullptr);

  const DepthwiseConv2DOptions* schema_params =
      op->builtin_options_as_DepthwiseConv2DOptions();

  if (schema_params != nullptr) {
    params->padding = ConvertPadding(schema_params->padding());
    params->stride_width = schema_params->stride_w();
    params->stride_height = schema_params->stride_h();
    params->depth_multiplier = schema_params->depth_multiplier();
    params->activation =
        ConvertActivation(schema_params->fused_activation_function());

    params->dilation_width_factor = schema_params->dilation_w_factor();
    params->dilation_height_factor = schema_params->dilation_h_factor();
  }

  *builtin_data = params.release();
  return kTfLiteOk;
}

It's not required to check the op version here. When the new implementation reads an old model file where dilation factors are missing, it will use 1 as the default value, and the new kernel will work consistently with the old kernel.

Change kernel registration

The MutableOpResolver (defined in lite/mutable_op_resolver.h) provides a few functions to register op kernels. The minimum and maximum version are 1 by default:

void AddBuiltin(tflite::BuiltinOperator op, TfLiteRegistration* registration,
                int min_version = 1, int max_version = 1);
void AddCustom(const char* name, TfLiteRegistration* registration,
               int min_version = 1, int max_version = 1);

The built-in ops are registered in lite/kernels/register.cc. In this example, we implemented a new op kernel which can handle DepthwiseConv2D version 1 and 2, so we need to change this line:

AddBuiltin(BuiltinOperator_DEPTHWISE_CONV_2D, Register_DEPTHWISE_CONV_2D());

to:

AddBuiltin(BuiltinOperator_DEPTHWISE_CONV_2D, Register_DEPTHWISE_CONV_2D(),
             /* min_version = */ 1,
             /* max_version = */ 2);

Change TFLite op version

The next step is to make TFLite populate the minimum version that's required to execute the op. In this example, it means:

  • Populate version=1 when dilation factors are all 1.
  • Populate version=2 otherwise.

Modify GetBuiltinOperatorVersion function for the operator in lite/tools/versioning/op_version.cc by adding the new version to the case of DepthwiseConv2D:

case BuiltinOperator_DEPTHWISE_CONV_2D:
  auto depthwise_conv_params =
      reinterpret_cast<TfLiteDepthwiseConvParams*>(op_sig.builtin_data);
  TFLITE_DCHECK(depthwise_conv_params != nullptr);
  if (depthwise_conv_params->dilation_width_factor != 1 ||
       depthwise_conv_params->dilation_height_factor != 1) {
    return 2;
  }
  return 1;

Update the operator version map

The last step is to add the new version info into the operator version map. This step is required because we need to generate the model's minimum required runtime version based on this version map.

To do this, you need to add a new map entry in lite/tools/versioning/runtime_version.cc.

In this example, you need to add the following entry into op_version_map:

{ {BuiltinOperator_DEPTHWISE_CONV_2D, 2}, %CURRENT_RUNTIME_VERSION%}

where %CURRENT_RUNTIME_VERSION% corresponds to the current runtime version defined in tensorflow/core/public/version.h.

Delegation implementation

TensorFlow Lite provides a delegation API which enables delegating ops to hardware backends. In the delegate's Prepare function, check if the version is supported for every node in Delegation code.

const int kMaxVersion = 1;
TfLiteNode* node;
TfLiteRegistration* registration = nullptr;
TF_LITE_ENSURE_STATUS(context->GetNodeAndRegistration(context, node_index, &node, &registration));

if (registration->version > kMaxVersion) {
  // Reject the node if the version isn't supported.
}

This is required even if the delegation only supports version 1 ops, so the delegation can detect incompatibility when getting a higher version op.