xref: /aosp_15_r20/external/tensorflow/tensorflow/core/kernels/data/padded_batch_dataset_op.cc (revision b6fb3261f9314811a0f4371741dbb8839866f948)

/* Copyright 2017 The TensorFlow Authors. All Rights Reserved.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#include "tensorflow/core/kernels/data/padded_batch_dataset_op.h"

#include "tensorflow/core/data/dataset_utils.h"
#include "tensorflow/core/data/name_utils.h"
#include "tensorflow/core/framework/dataset.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/partial_tensor_shape.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/framework/tensor_util.h"
#include "tensorflow/core/lib/core/blocking_counter.h"
#include "tensorflow/core/lib/core/errors.h"
#include "tensorflow/core/lib/gtl/cleanup.h"
#include "tensorflow/core/platform/macros.h"
#include "tensorflow/core/platform/stringprintf.h"
#include "tensorflow/core/util/batch_util.h"

namespace tensorflow {
namespace data {

// See documentation in ../../ops/dataset_ops.cc for a high-level
// description of the following op.

/* static */ constexpr const char* const PaddedBatchDatasetOp::kDatasetType;
/* static */ constexpr const char* const PaddedBatchDatasetOp::kInputDataset;
/* static */ constexpr const char* const PaddedBatchDatasetOp::kBatchSize;
/* static */ constexpr const char* const PaddedBatchDatasetOp::kPaddedShapes;
/* static */ constexpr const char* const PaddedBatchDatasetOp::kPaddingValues;
/* static */ constexpr const char* const PaddedBatchDatasetOp::kDropRemainder;
/* static */ constexpr const char* const PaddedBatchDatasetOp::kParallelCopy;
/* static */ constexpr const char* const PaddedBatchDatasetOp::kToutputTypes;
/* static */ constexpr const char* const PaddedBatchDatasetOp::kOutputShapes;
/* static */ constexpr const char* const PaddedBatchDatasetOp::kNumPaddedShapes;

constexpr char kExhausted[] = "exhausted";

class PaddedBatchDatasetOp::Dataset : public DatasetBase {
 public:
  Dataset(OpKernelContext* ctx, int64_t batch_size, bool drop_remainder,
          bool parallel_copy, std::vector<PartialTensorShape> padded_shapes,
          std::vector<Tensor> padding_values, const DatasetBase* input,
          int op_version)
      : DatasetBase(DatasetContext(ctx)),
        batch_size_(batch_size),
        drop_remainder_(drop_remainder),
        parallel_copy_(parallel_copy),
        padded_shapes_(std::move(padded_shapes)),
        padding_values_(std::move(padding_values)),
        input_(input),
        op_version_(op_version),
        traceme_metadata_(
            {{"batch_size",
              strings::Printf("%lld", static_cast<long long>(batch_size))},
             {"drop_remainder", drop_remainder ? "true" : "false"},
             {"parallel_copy", parallel_copy ? "true" : "false"}}) {
    input_->Ref();

    // NOTE(mrry): Currently we implement "batch up to" semantics. If we could
    // tell statically that the input dataset is infinite, then we could
    // always report `batch_size` as the 0th dimension.
    //
    // TODO(mrry): Need to validate that the input shape and the padded shape
    // are "compatible" (i.e. that padded shape is >= input shape, with both
    // static and dynamic checks as appropriate).
    const auto& input_shapes = input_->output_shapes();
    output_shapes_.reserve(input_shapes.size());
    for (size_t i = 0; i < input_shapes.size(); ++i) {
      if (drop_remainder_ || input_->Cardinality() == kInfiniteCardinality) {
        output_shapes_.push_back(
            PartialTensorShape({batch_size_}).Concatenate(padded_shapes_[i]));
      } else {
        output_shapes_.push_back(
            PartialTensorShape({-1}).Concatenate(padded_shapes_[i]));
      }
    }
  }

  ~Dataset() override { input_->Unref(); }

  std::unique_ptr<IteratorBase> MakeIteratorInternal(
      const string& prefix) const override {
    name_utils::IteratorPrefixParams params;
    params.op_version = op_version_;
    return std::make_unique<Iterator>(Iterator::Params{
        this, name_utils::IteratorPrefix(kDatasetType, prefix, params)});
  }

  const DataTypeVector& output_dtypes() const override {
    return input_->output_dtypes();
  }

  const std::vector<PartialTensorShape>& output_shapes() const override {
    return output_shapes_;
  }

  string DebugString() const override {
    name_utils::DatasetDebugStringParams params;
    params.op_version = op_version_;
    params.set_args(batch_size_);
    return name_utils::DatasetDebugString(kDatasetType, params);
  }

  int64_t CardinalityInternal() const override {
    int64_t n = input_->Cardinality();
    if (n == kInfiniteCardinality || n == kUnknownCardinality) {
      return n;
    }
    return n / batch_size_ + (n % batch_size_ == 0 || drop_remainder_ ? 0 : 1);
  }

  Status InputDatasets(std::vector<const DatasetBase*>* inputs) const override {
    inputs->push_back(input_);
    return OkStatus();
  }

  Status CheckExternalState() const override {
    return input_->CheckExternalState();
  }

 protected:
  Status AsGraphDefInternal(SerializationContext* ctx,
                            DatasetGraphDefBuilder* b,
                            Node** output) const override {
    Node* input_graph_node = nullptr;
    TF_RETURN_IF_ERROR(b->AddInputDataset(ctx, input_, &input_graph_node));
    Node* batch_size = nullptr;
    TF_RETURN_IF_ERROR(b->AddScalar(batch_size_, &batch_size));

    std::vector<Node*> padded_shapes;
    padded_shapes.reserve(padded_shapes_.size());
    for (int i = 0; i < padded_shapes_.size(); i++) {
      Node* node;
      Tensor t(DT_INT64, TensorShape({padded_shapes_[i].dims()}));
      for (int j = 0; j < padded_shapes_[i].dims(); j++) {
        t.vec<int64_t>()(j) = padded_shapes_[i].dim_size(j);
      }
      TF_RETURN_IF_ERROR(b->AddTensor(t, &node));
      padded_shapes.emplace_back(node);
    }

    std::vector<Node*> padding_values;
    padding_values.reserve(padding_values_.size());
    for (const Tensor& t : padding_values_) {
      Node* node;
      TF_RETURN_IF_ERROR(b->AddTensor(t, &node));
      padding_values.emplace_back(node);
    }

    Node* drop_remainder = nullptr;
    TF_RETURN_IF_ERROR(b->AddScalar(drop_remainder_, &drop_remainder));

    AttrValue parallel_copy;
    b->BuildAttrValue(parallel_copy_, &parallel_copy);

    AttrValue output_types;
    b->BuildAttrValue(output_dtypes(), &output_types);

    AttrValue N;
    b->BuildAttrValue<int64_t>(padded_shapes_.size(), &N);

    TF_RETURN_IF_ERROR(b->AddDataset(
        this, {{0, input_graph_node}, {1, batch_size}, {4, drop_remainder}},
        {{2, padded_shapes}, {3, padding_values}},
        {{kParallelCopy, parallel_copy},
         {kToutputTypes, output_types},
         {kNumPaddedShapes, N}},
        output));
    return OkStatus();
  }

 private:
  class Iterator : public DatasetIterator<Dataset> {
   public:
    explicit Iterator(const Params& params)
        : DatasetIterator<Dataset>(params) {}

    Status Initialize(IteratorContext* ctx) override {
      return dataset()->input_->MakeIterator(ctx, this, prefix(), &input_impl_);
    }

    Status GetNextInternal(IteratorContext* ctx,
                           std::vector<Tensor>* out_tensors,
                           bool* end_of_sequence) override {
      // Each row of `batch_elements` is a tuple of tensors from the
      // input iterator.
      std::vector<std::vector<Tensor>> batch_elements;
      {
        mutex_lock l(mu_);
        if (!input_impl_) {
          *end_of_sequence = true;
          return OkStatus();
        } else {
          *end_of_sequence = false;
          batch_elements.reserve(dataset()->batch_size_);
          for (int i = 0; i < dataset()->batch_size_ && !*end_of_sequence;
               ++i) {
            std::vector<Tensor> batch_element_tuple;
            TF_RETURN_IF_ERROR(input_impl_->GetNext(ctx, &batch_element_tuple,
                                                    end_of_sequence));
            if (!*end_of_sequence) {
              batch_elements.push_back(std::move(batch_element_tuple));
            }
          }
          if (*end_of_sequence) {
            input_impl_.reset();
          }
        }
      }

      if (batch_elements.empty()) {
        DCHECK(*end_of_sequence);
        return OkStatus();
      }

      if (dataset()->drop_remainder_ &&
          batch_elements.size() < dataset()->batch_size_) {
        *end_of_sequence = true;
        return OkStatus();
      }

      TF_RETURN_IF_ERROR(CopyBatch(ctx, batch_elements, out_tensors));
      *end_of_sequence = false;
      return OkStatus();
    }

   protected:
    std::shared_ptr<model::Node> CreateNode(
        IteratorContext* ctx, model::Node::Args args) const override {
      return model::MakeKnownRatioNode(std::move(args), dataset()->batch_size_);
    }

    Status SaveInternal(SerializationContext* ctx,
                        IteratorStateWriter* writer) override {
      mutex_lock l(mu_);
      if (input_impl_)
        TF_RETURN_IF_ERROR(SaveInput(ctx, writer, input_impl_));
      else
        TF_RETURN_IF_ERROR(writer->WriteScalar(full_name(kExhausted), ""));
      return OkStatus();
    }

    Status RestoreInternal(IteratorContext* ctx,
                           IteratorStateReader* reader) override {
      mutex_lock l(mu_);
      if (reader->Contains(full_name(kExhausted))) {
        input_impl_.reset();
      } else {
        TF_RETURN_IF_ERROR(
            dataset()->input_->MakeIterator(ctx, this, prefix(), &input_impl_));
        TF_RETURN_IF_ERROR(RestoreInput(ctx, reader, input_impl_));
      }
      return OkStatus();
    }

    TraceMeMetadata GetTraceMeMetadata() const override {
      return dataset()->traceme_metadata_;
    }

   private:
    // Copies the retrieved batch elements into one output tensor per tuple
    // component.
    //
    // NOTE(mrry): If the input or output sizes are statically known, we could
    // potentially read the input values in-place into their respective slice
    // locations. This would require a different GetNext() overload that
    // supports zero-copy, and might make sense in an optimization pass.
    Status CopyBatch(IteratorContext* ctx,
                     const std::vector<std::vector<Tensor>>& batch_elements,
                     std::vector<Tensor>* out_tensors) {
      const size_t num_tuple_components = batch_elements[0].size();
      const int64_t num_batch_elements = batch_elements.size();
      for (size_t component_index = 0; component_index < num_tuple_components;
           ++component_index) {
        // 1. Determine the shape of the padded tensor.
        TensorShape batch_component_shape({num_batch_elements});
        const PartialTensorShape& padded_shape =
            dataset()->padded_shapes_[component_index];

        for (int dim = 0; dim < padded_shape.dims(); ++dim) {
          if (padded_shape.dim_size(dim) == -1) {
            batch_component_shape.AddDim(0);
          } else {
            batch_component_shape.AddDim(padded_shape.dim_size(dim));
          }
        }

        for (int64_t i = 0; i < num_batch_elements; ++i) {
          const TensorShape& element_shape =
              batch_elements[i][component_index].shape();
          // TODO(mrry): Perform this check in the shape function if
          // enough static information is available to do so.
          if (element_shape.dims() != padded_shape.dims()) {
            return errors::InvalidArgument(
                "All elements in a batch must have the same rank as the "
                "padded shape for component",
                component_index, ": expected rank ", padded_shape.dims(),
                " but got element with rank ", element_shape.dims());
          }
          for (int dim = 0; dim < padded_shape.dims(); ++dim) {
            if (padded_shape.dim_size(dim) == -1) {
              // Take the max of all batch elements in this dimension.
              if (batch_elements[i][component_index].shape().dim_size(dim) >
                  batch_component_shape.dim_size(dim + 1)) {
                batch_component_shape.set_dim(
                    dim + 1,
                    batch_elements[i][component_index].shape().dim_size(dim));
              }
            } else {
              if (batch_elements[i][component_index].shape().dim_size(dim) >
                  batch_component_shape.dim_size(dim + 1)) {
                return errors::DataLoss(
                    "Attempted to pad to a smaller size than the input "
                    "element.");
              }
            }
          }
        }

        // 2. Copy each batch element to the appropriate location in
        // the output component tensor.
        out_tensors->emplace_back(ctx->allocator({}),
                                  output_dtypes()[component_index],
                                  batch_component_shape);
        Tensor& batch_component = out_tensors->back();
        TF_RETURN_IF_ERROR(batch_util::SetElementZero(
            &batch_component, dataset()->padding_values_[component_index]));

        // Build the output tuple component by copying one slice from each input
        // element in the batch.
        TensorShape component_shape({});
        for (int i = 1; i < batch_component_shape.dims(); ++i) {
          component_shape.AddDim(batch_component_shape.dim_size(i));
        }
        auto copy_element_fn = [component_index, &batch_elements,
                                &batch_component, &component_shape](int index) {
          // Take the fast path if possible.
          if (batch_elements[index][component_index].shape() ==
              component_shape) {
            TF_RETURN_IF_ERROR(batch_util::CopyElementToSlice(
                batch_elements[index][component_index], &batch_component,
                index));
          } else {
            TF_RETURN_IF_ERROR(batch_util::CopyElementToLargerSlice(
                batch_elements[index][component_index], &batch_component,
                index));
          }
          return OkStatus();
        };

        if (dataset()->parallel_copy_ && (batch_component.AllocatedBytes() /
                                          num_batch_elements) >= (1 << 15)) {
          BlockingCounter counter(num_batch_elements);
          Status status;
          mutex status_mu;
          const auto num_threads = ctx->runner_threadpool_size();
          const auto slice_size = num_batch_elements / num_threads;
          int64_t offset = 0;
          for (size_t i = 0; i < num_threads; ++i) {
            int64_t length = slice_size;
            // When the number of threads does not divide the number of elements
            // evenly, the size of some slices is incremented to guarantee their
            // sizes add up to the total number of elements.
            if (i < num_batch_elements % num_threads) ++length;
            (*ctx->runner())([offset, length, &status, &status_mu, &counter,
                              &copy_element_fn]() {
              for (size_t j = offset; j < offset + length; ++j) {
                {
                  Status s = copy_element_fn(j);
                  mutex_lock l(status_mu);
                  status.Update(s);
                }
                counter.DecrementCount();
              }
            });
            offset += length;
          }
          counter.Wait();
          TF_RETURN_IF_ERROR(status);
        } else {
          for (size_t i = 0; i < num_batch_elements; ++i) {
            TF_RETURN_IF_ERROR(copy_element_fn(i));
          }
        }
      }
      return OkStatus();
    }

    mutex mu_;
    std::unique_ptr<IteratorBase> input_impl_ TF_GUARDED_BY(mu_);
  };

  const int64_t batch_size_;
  const bool drop_remainder_;
  const bool parallel_copy_;
  const std::vector<PartialTensorShape> padded_shapes_;
  const std::vector<Tensor> padding_values_;
  const DatasetBase* const input_;
  const int op_version_;
  std::vector<PartialTensorShape> output_shapes_;
  const TraceMeMetadata traceme_metadata_;
};

PaddedBatchDatasetOp::PaddedBatchDatasetOp(OpKernelConstruction* ctx)
    : UnaryDatasetOpKernel(ctx),
      op_version_(ctx->def().op() == "PaddedBatchDataset" ? 1 : 2) {
  if (ctx->HasAttr(kParallelCopy)) {
    OP_REQUIRES_OK(ctx, ctx->GetAttr(kParallelCopy, &parallel_copy_));
  }
}

void PaddedBatchDatasetOp::MakeDataset(OpKernelContext* ctx, DatasetBase* input,
                                       DatasetBase** output) {
  int64_t batch_size;
  OP_REQUIRES_OK(ctx,
                 ParseScalarArgument<int64_t>(ctx, kBatchSize, &batch_size));
  OP_REQUIRES(ctx, batch_size > 0,
              errors::InvalidArgument("Batch size must be greater than zero."));

  bool drop_remainder = false;
  if (op_version_ > 1) {
    OP_REQUIRES_OK(
        ctx, ParseScalarArgument<bool>(ctx, kDropRemainder, &drop_remainder));
  }

  OpInputList padded_shape_tensors;
  OP_REQUIRES_OK(ctx, ctx->input_list(kPaddedShapes, &padded_shape_tensors));
  std::vector<PartialTensorShape> padded_shapes;
  padded_shapes.reserve(padded_shape_tensors.size());
  OP_REQUIRES(ctx, padded_shape_tensors.size() == input->output_shapes().size(),
              errors::InvalidArgument("Number of padded shapes (",
                                      padded_shape_tensors.size(),
                                      ") must match the number of components "
                                      "in the input dataset's elements (",
                                      input->output_shapes().size(), ")"));
  for (const Tensor& padded_shape_t : padded_shape_tensors) {
    OP_REQUIRES(ctx, TensorShapeUtils::IsVector(padded_shape_t.shape()),
                errors::InvalidArgument("All padded shapes must be vectors"));
    PartialTensorShape padded_shape;
    OP_REQUIRES_OK(ctx, PartialTensorShape::MakePartialShape(
                            padded_shape_t.vec<int64_t>().data(),
                            padded_shape_t.NumElements(), &padded_shape));
    padded_shapes.push_back(std::move(padded_shape));
  }
  OpInputList padding_values_list;
  OP_REQUIRES_OK(ctx, ctx->input_list(kPaddingValues, &padding_values_list));
  std::vector<Tensor> padding_values;
  OP_REQUIRES(ctx, padding_values_list.size() == input->output_shapes().size(),
              errors::InvalidArgument(
                  "Number of padding values (", padding_values_list.size(),
                  ") must match the number of components in the input "
                  "dataset's elements (",
                  input->output_shapes().size(), ")"));
  for (int i = 0; i < padding_values_list.size(); ++i) {
    const Tensor& padding_value_t = padding_values_list[i];
    OP_REQUIRES(ctx, TensorShapeUtils::IsScalar(padding_value_t.shape()),
                errors::InvalidArgument("All padding values must be scalars"));
    OP_REQUIRES(ctx, padding_value_t.dtype() == input->output_dtypes()[i],
                errors::InvalidArgument(
                    "Mismatched type between padding value ", i,
                    " and input dataset's component ", i, ": ",
                    DataTypeString(padding_value_t.dtype()), " vs. ",
                    DataTypeString(input->output_dtypes()[i])));
    padding_values.push_back(tensor::DeepCopy(padding_value_t));
  }

  *output = new Dataset(ctx, batch_size, drop_remainder, parallel_copy_,
                        std::move(padded_shapes), std::move(padding_values),
                        input, op_version_);
}

namespace {
REGISTER_KERNEL_BUILDER(Name("PaddedBatchDataset").Device(DEVICE_CPU),
                        PaddedBatchDatasetOp);

REGISTER_KERNEL_BUILDER(Name("PaddedBatchDatasetV2").Device(DEVICE_CPU),
                        PaddedBatchDatasetOp);
}  // namespace
}  // namespace data
}  // namespace tensorflow