parameterized_truncated_normal_op_gpu.cu.cc

/* Copyright 2015 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.
==============================================================================*/

#if GOOGLE_CUDA || TENSORFLOW_USE_ROCM

#define EIGEN_USE_GPU

#include <assert.h>
#include <stdio.h>

#include <cmath>

#include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
#include "tensorflow/core/framework/tensor_types.h"
#include "tensorflow/core/kernels/parameterized_truncated_normal_op.h"
#include "tensorflow/core/lib/random/philox_random.h"
#include "tensorflow/core/lib/random/random_distributions.h"
#include "tensorflow/core/util/gpu_kernel_helper.h"

#if defined(_MSC_VER) && !defined(__clang__)
// msvc does not support unroll. One could try the loop pragma but we need to
// take a closer look if this generates better code in this case. For now let
// the compiler take care of it.
#define UNROLL
#else
#define UNROLL _Pragma("unroll")
#endif

namespace tensorflow {

class OpKernelContext;

namespace functor {

static constexpr int kMaxIterations = 1000;

typedef Eigen::GpuDevice GPUDevice;

template <typename T>

__global__ void __launch_bounds__(1024)
    TruncatedNormalKernel(random::PhiloxRandom gen, T* data, int64 num_batches,
                          int64 samples_per_batch, int64 num_elements,
                          const T* __restrict__ means, bool single_mean,
                          const T* __restrict__ stddevs, bool single_stddev,
                          const T* __restrict__ minvals, bool single_minval,
                          const T* __restrict__ maxvals, bool single_maxval,
                          int64 kMaxIterations) {
  const int32 max_samples_per_item = 2 * kMaxIterations;
  // Initial offset as given by GPU_1D_KERNEL_LOOP.
  const int32 initial_offset = blockIdx.x * blockDim.x + threadIdx.x;
  gen.Skip(max_samples_per_item * initial_offset);
  typedef random::UniformDistribution<random::PhiloxRandom, T> Uniform;
  typedef random::NormalDistribution<random::PhiloxRandom, T> Normal;
  Uniform dist;
  Normal normal_dist;
  const int kDistSize = Uniform::kResultElementCount;
  const T quietNaN = Eigen::NumTraits<T>::quiet_NaN();
  // The randn rejection sampling is used when the mean and at least this many
  // standard deviations are inside the bounds.
  // The uniform proposal samplers become less efficient as the bounds are
  // further from the mean, the reverse is true for the randn sampler.
  // This number was chosen by empirical benchmarking. If modified, the
  // benchmarks in parameterized_truncated_normal_op_test should also be
  // changed.
  const T kStdDevsInsideBoundsToUseRandnSampler = T(1.7);

  // We skip the total number of threads to get to the next element. To produce
  // deterministic results between devices, each element in the output array
  // skips max_samples_per_item in the generator. Then after generating this
  // item, we need to skip the samples for one element for every thread to get
  // to the next element that we actually process.
  const int32 samples_between_processed_elements =
      max_samples_per_item * (gridDim.x * blockDim.x);

  GPU_1D_KERNEL_LOOP(offset, num_elements) {
    // Track how many more samples we need to skip before we process the next
    // element.
    int32 remaining_samples = samples_between_processed_elements;

    const int64 batch_id = offset / samples_per_batch;
    T mean = means[single_mean ? 0 : batch_id];
    const T input_stddev = stddevs[single_stddev ? 0 : batch_id];
    T minval = minvals[single_minval ? 0 : batch_id];
    T maxval = maxvals[single_maxval ? 0 : batch_id];

    // Flip the distribution if we can make the lower bound positive.
    T stddev;
    if (Eigen::numext::isinf(minval) || maxval < mean) {
      // Reverse all calculations. normMin and normMax will be flipped.
      // std::swap is a host function (not available in CUDA).
      T temp = minval;
      minval = maxval;
      maxval = temp;
      stddev = -input_stddev;
    } else {
      stddev = input_stddev;
    }

    // Calculate normalized samples, then scale them.
    const T normMin = (minval - mean) / stddev;
    const T normMax = (maxval - mean) / stddev;

    // Determine the method to use.
    const T sqrtFactor = Eigen::numext::sqrt((normMin * normMin) + T(4));
    const T cutoff =
        T(2) *
        Eigen::numext::exp(T(0.5) + (normMin * (normMin - sqrtFactor)) / T(4)) /
        (normMin + sqrtFactor);
    const T diff = normMax - normMin;
    const T two = T(2.0);

    // Validate the normalized min and max, because the originals may have been
    // flipped already.
    if (!(input_stddev > T(0) && normMin < normMax &&
          (Eigen::numext::isfinite(normMin) ||
           Eigen::numext::isfinite(normMax)))) {
      data[offset] = quietNaN;
    } else if (((normMin < -kStdDevsInsideBoundsToUseRandnSampler) &&
                (normMax >= T(0.))) ||
               ((normMax > kStdDevsInsideBoundsToUseRandnSampler) &&
                (normMin <= T(0.)))) {
      Eigen::array<T, 4> n;

      int numIterations = 0;
      while (numIterations < kMaxIterations) {
        const auto randn = normal_dist(&gen);
        remaining_samples -= gen.kResultElementCount;
        UNROLL for (int i = 0; i < kDistSize; i++) {
          if ((randn[i] >= normMin) && randn[i] <= normMax) {
            data[offset] = randn[i] * stddev + mean;
            numIterations = kMaxIterations;
            break;
          } else if (numIterations + 1 == kMaxIterations) {
            // If we did not successfully sample after all these iterations
            // something is wrong. Output a nan.
            data[offset] = quietNaN;
            numIterations = kMaxIterations;
            break;
          } else {
            numIterations++;
          }
        }
      }
    } else if (diff < cutoff) {
      // Sample from a uniform distribution on [normMin, normMax].

      // Vectorized intermediate calculations for uniform rejection sampling.
      // We always generate at most 4 samples.
      Eigen::array<T, 4> z;
      Eigen::array<T, 4> g;

      const T plusFactor = (normMin < T(0)) ? T(0) : T(normMin * normMin);

      int numIterations = 0;
      while (numIterations < kMaxIterations) {
        const auto rand = dist(&gen);
        remaining_samples -= gen.kResultElementCount;
        UNROLL for (int i = 0; i < kDistSize; i++) {
          z[i] = rand[i] * diff + normMin;
        }
        UNROLL for (int i = 0; i < kDistSize; i++) {
          g[i] = (plusFactor - z[i] * z[i]) / two;
        }

        const auto u = dist(&gen);
        remaining_samples -= gen.kResultElementCount;
        UNROLL for (int i = 0; i < kDistSize; i++) {
          bool accept = u[i] <= Eigen::numext::exp(g[i]);
          if (accept) {
            // Accept the sample z.
            data[offset] = z[i] * stddev + mean;
            // Break out of the nested loop by updating numIterations.
            numIterations = kMaxIterations;
            break;
          } else if (numIterations + 1 >= kMaxIterations) {
            data[offset] = quietNaN;
            numIterations = kMaxIterations;
            break;
          } else {
            numIterations++;
          }
        }
      }
    } else {
      // Sample from an exponential distribution with alpha maximizing
      // acceptance probability, offset by normMin from the origin.
      // Accept only if less than normMax.
      const T alpha =
          (normMin + Eigen::numext::sqrt((normMin * normMin) + T(4))) / T(2);
      int numIterations = 0;
      while (numIterations < kMaxIterations) {
        auto rand = dist(&gen);
        remaining_samples -= gen.kResultElementCount;
        UNROLL for (int i = 0; i < kDistSize; i += 2) {
          const T z = -Eigen::numext::log(rand[i]) / alpha + normMin;
          const T x = normMin < alpha ? alpha - z : normMin - alpha;
          const T g = Eigen::numext::exp(-x * x / two);
          const T u = rand[i + 1];
          bool accept = (u <= g && z < normMax);
          if (accept) {
            data[offset] = z * stddev + mean;
            // Break out of the nested loop by updating numIterations.
            numIterations = kMaxIterations;
            break;
          } else if (numIterations + 1 >= kMaxIterations) {
            data[offset] = quietNaN;
            numIterations = kMaxIterations;
            break;
          } else {
            numIterations++;
          }
        }
      }
    }

    gen.Skip(remaining_samples);
  }
}

// Partial specialization for GPU
template <typename T>
struct TruncatedNormalFunctor<GPUDevice, T> {
  void operator()(OpKernelContext* ctx, const GPUDevice& d, int64 num_batches,
                  int64 samples_per_batch, int64 num_elements,
                  typename TTypes<T>::ConstFlat means,
                  typename TTypes<T>::ConstFlat stddevs,
                  typename TTypes<T>::ConstFlat minvals,
                  typename TTypes<T>::ConstFlat maxvals,
                  const random::PhiloxRandom& gen,
                  typename TTypes<T>::Flat output) {
    const auto config = GetGpuLaunchConfig(num_elements, d);

    TF_CHECK_OK(GpuLaunchKernel(
        TruncatedNormalKernel<T>, config.block_count, config.thread_per_block,
        0, d.stream(), gen, output.data(), num_batches, samples_per_batch,
        num_elements, means.data(), means.dimension(0) == 1, stddevs.data(),
        stddevs.dimension(0) == 1, minvals.data(), minvals.dimension(0) == 1,
        maxvals.data(), maxvals.dimension(0) == 1, kMaxIterations));
  }
};

// Explicit instantiation of the GPU distributions functors
template struct TruncatedNormalFunctor<GPUDevice, Eigen::half>;
template struct TruncatedNormalFunctor<GPUDevice, float>;
template struct TruncatedNormalFunctor<GPUDevice, double>;

}  // namespace functor
}  // namespace tensorflow

#endif  // GOOGLE_CUDA || TENSORFLOW_USE_ROCM