multihead_matmul_op.cu

// Copyright (c) 2019 PaddlePaddle 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 <paddle/fluid/platform/device_context.h>

#include <algorithm>
#include <type_traits>

#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/memory/malloc.h"
#include "paddle/fluid/operators/math/bert_encoder_functor.h"
#include "paddle/fluid/platform/float16.h"
#include "paddle/phi/kernels/funcs/blas/blas.h"

namespace paddle {
namespace operators {

template <typename T>
__global__ void transpose(T *src,
                          T *dst,
                          const int batch_size,
                          const int seq_len,
                          const int head_num,
                          const int size_per_head) {
  int batch_id = blockIdx.x / (head_num * seq_len);
  int seq_id = blockIdx.x % seq_len;
  int head_id = (blockIdx.x % (head_num * seq_len)) / seq_len;
  dst[batch_id * (head_num * seq_len * size_per_head) +
      seq_id * head_num * size_per_head + head_id * size_per_head +
      threadIdx.x] = src[blockIdx.x * size_per_head + threadIdx.x];
}

template <typename T>
inline __device__ T add_func(T a, T b);

template <>
__device__ float add_func<float>(float a, float b) {
  return a + b;
}

template <>
__device__ float2 add_func<float2>(float2 a, float2 b) {
  float2 c;
  c.x = a.x + b.x;
  c.y = a.y + b.y;
  return c;
}

template <>
__device__ float4 add_func<float4>(float4 a, float4 b) {
  float4 c;
  c.x = a.x + b.x;
  c.y = a.y + b.y;
  c.z = a.z + b.z;
  c.w = a.w + b.w;
  return c;
}
#if defined(PADDLE_WITH_CUDA)
template <>
__device__ half2 add_func<half2>(half2 a, half2 b) {
#if __CUDA_ARCH__ >= 530
  return __hadd2(a, b);
#else
  return half2(__float2half(__half2float(a.x) + __half2float(b.x)),
               __float2half(__half2float(b.x) + __half2float(b.y)));
#endif
}

template <>
__device__ half add_func<half>(half a, half b) {
#if __CUDA_ARCH__ >= 530
  return __hadd(a, b);
#else
  return __float2half(__half2float(a) + __half2float(b));
#endif
}
#endif

template <typename T>
__global__ void TransposeQkvKernel(const int H,
                                   const T *input,
                                   const T *bias,
                                   T *output) {
  // Input: BxSx3xNxH
  // Bias: 3xNxH
  // Output: 3xBxNxSxH
  int n = threadIdx.y;
  int s = blockIdx.x;
  int b = blockIdx.y;
  int m = blockIdx.z;

  const int N = blockDim.y;
  const int S = gridDim.x;
  const int B = gridDim.y;

  const int NH = N * H;
  const int NHS = NH * S;
  const int in_offset = n * H + m * NH + s * 3 * NH + b * NHS * 3;
  const int bias_offset = m * NH + n * H;
  const int out_offset = s * H + n * S * H + b * NHS + m * NHS * B;

  const int i = threadIdx.x;
  output[out_offset + i] =
      add_func(input[in_offset + i], bias[bias_offset + i]);
}

template <typename T>
void TransQKVWithBias(const int batch,
                      const int seq_len,
                      const int head_size,
                      const int head_num,
                      const T *input,
                      const T *bias,
                      T *output,
                      gpuStream_t stream);

template <>
void TransQKVWithBias(const int batch,
                      const int seq_len,
                      const int head_size,
                      const int head_num,
                      const float *input,
                      const float *bias,
                      float *output,
                      gpuStream_t stream) {
  // BxSx3xNxH + 3xNxH -> 3xBxNxSxH
  int scratch_size = batch * head_num * seq_len * seq_len;
  const dim3 grid(seq_len, batch, 3);
  // scratch % 4 == 0 to ensure the alignment
  if (head_size % 4 == 0 && scratch_size % 4 == 0) {
    const int h = head_size / 4;
    const float4 *input4 = reinterpret_cast<const float4 *>(input);
    const float4 *bias4 = reinterpret_cast<const float4 *>(bias);
    float4 *output4 = reinterpret_cast<float4 *>(output);
    const dim3 block(h, head_num, 1);

    // limit h * head_num to max block size(1024).
    PADDLE_ENFORCE_LE(h * head_num,
                      1024,
                      platform::errors::InvalidArgument(
                          "head_num (%d) * head_size (%d) should <= %d",
                          head_num,
                          head_size,
                          1024 * 4));
    TransposeQkvKernel<float4>
        <<<grid, block, 0, stream>>>(h, input4, bias4, output4);
  } else if (head_size % 2 == 0 && scratch_size % 2 == 0) {
    const int h = head_size / 2;
    const float2 *input2 = reinterpret_cast<const float2 *>(input);
    const float2 *bias2 = reinterpret_cast<const float2 *>(bias);
    float2 *output2 = reinterpret_cast<float2 *>(output);
    const dim3 block(h, head_num, 1);
    // limit h * head_num to max block size(1024).
    PADDLE_ENFORCE_LE(h * head_num,
                      1024,
                      platform::errors::InvalidArgument(
                          "head_num (%d) * head_size (%d) should <= %d",
                          head_num,
                          head_size,
                          1024 * 2));
    TransposeQkvKernel<float2>
        <<<grid, block, 0, stream>>>(h, input2, bias2, output2);
  } else {
    const dim3 block(head_size, head_num, 1);
    // limit head_size * head_num to max block size(1024).
    PADDLE_ENFORCE_LE(head_size * head_num,
                      1024,
                      platform::errors::InvalidArgument(
                          "head_num (%d) * head_size (%d) should <= %d",
                          head_num,
                          head_size,
                          1024));
    TransposeQkvKernel<float>
        <<<grid, block, 0, stream>>>(head_size, input, bias, output);
  }
}

#if defined(PADDLE_WITH_CUDA)
template <>
void TransQKVWithBias(const int batch,
                      const int seq_len,
                      const int head_size,
                      const int head_num,
                      const platform::float16 *input,
                      const platform::float16 *bias,
                      platform::float16 *output,
                      gpuStream_t stream) {
  // BxSx3xNxH + 3xNxH -> 3xBxNxSxH
  int scratch_size = batch * head_num * seq_len * seq_len;
  const dim3 grid(seq_len, batch, 3);
  if (head_size % 2 == 0 && scratch_size % 2 == 0) {
    const int h = head_size / 2;
    const half2 *input2 = reinterpret_cast<const half2 *>(input);
    const half2 *bias2 = reinterpret_cast<const half2 *>(bias);
    half2 *output2 = reinterpret_cast<half2 *>(output);
    const dim3 block(h, head_num, 1);
    // limit h * head_num to max block size(1024).
    PADDLE_ENFORCE_LE(h * head_num,
                      1024,
                      platform::errors::InvalidArgument(
                          "head_num (%d) * head_size (%d) should <= %d",
                          head_num,
                          head_size,
                          1024 * 2));
    TransposeQkvKernel<half2>
        <<<grid, block, 0, stream>>>(h, input2, bias2, output2);
  } else {
    const dim3 block(head_size, head_num, 1);
    const half *input_half = reinterpret_cast<const half *>(input);
    const half *bias_half = reinterpret_cast<const half *>(bias);
    half *output_half = reinterpret_cast<half *>(output);

    // limit head_size * head_num to max block size(1024).
    PADDLE_ENFORCE_LE(head_size * head_num,
                      1024,
                      platform::errors::InvalidArgument(
                          "head_num (%d) * head_size (%d) should <= %d",
                          head_num,
                          head_size,
                          1024));
    TransposeQkvKernel<half><<<grid, block, 0, stream>>>(
        head_size, input_half, bias_half, output_half);
  }
}
#endif

inline int round_up(int seq_len, int multiple = 32) {
  PADDLE_ENFORCE_GT(
      multiple,
      0,
      platform::errors::InvalidArgument(
          "multiple should be a positive number，but it's (%d)", multiple));
  return ((seq_len + multiple - 1) / multiple) * multiple;
}

template <typename T>
__global__ void broadcast(const T *src,
                          T *dst,
                          const int seq_len,
                          const int head_num) {
  int batch_id = blockIdx.x / (head_num * seq_len);
  int dst_offset = blockIdx.x * seq_len;
  if (threadIdx.x < seq_len) {
    dst[threadIdx.x + dst_offset] = src[threadIdx.x + batch_id * seq_len];
  }
}

template <typename DeviceContext, typename T>
class MultiHeadMatMulV2Kernel : public framework::OpKernel<T> {
 public:
  void Compute(const framework::ExecutionContext &context) const override {
    using Tensor = framework::Tensor;
    auto *input = context.Input<framework::Tensor>("Input");
    auto *w = context.Input<framework::Tensor>("W");
    auto *bias = context.Input<framework::Tensor>("Bias");
    auto &bias_qk = GET_DATA_SAFELY(context.Input<framework::Tensor>("BiasQK"),
                                    "Input",
                                    "BiasQK",
                                    "MultiHeadMatMulV2");

    auto *input_d = input->data<T>();
    auto *w_d = w->data<T>();
    auto *bias_d = bias->data<T>();
    auto *bias_qk_d = bias_qk.template data<T>();
    T scale = static_cast<T>(context.Attr<float>("alpha"));

    int head_number = context.Attr<int>("head_number");
    // compute q*k with eltadd
    auto &device_ctx = context.template device_context<DeviceContext>();
    auto stream = device_ctx.stream();
    // should be (B * S * hidden)
    auto input_dims = input->dims();
    // shouble be (hidden * 3 * all_head_size)
    auto w_dims = w->dims();
    int batch = input_dims[0];
    int seq_len = input_dims[1];
    int hidden = input_dims[2];
    Tensor temp_bias_tensor;
    // if bias_qk is[batch, 1, 1, seq_len], the bias_qk_d need to be broadcasted
    if (bias_qk.numel() == (batch * seq_len)) {
      temp_bias_tensor.Resize({batch * head_number * seq_len * seq_len});
      auto *temp_qk_bias = temp_bias_tensor.mutable_data<T>(context.GetPlace());
      int grid = batch * head_number * seq_len;
      int block = round_up(seq_len);
      broadcast<<<grid, block, 0, stream>>>(
          bias_qk_d, temp_qk_bias, seq_len, head_number);
      bias_qk_d = static_cast<const T *>(temp_qk_bias);
    }
    int all_head_size = w_dims[2];
    int head_size = all_head_size / head_number;

    auto *out = context.Output<framework::Tensor>("Out");
    out->Resize({batch, seq_len, all_head_size});
    auto *output_d = out->mutable_data<T>(context.GetPlace());

    // (B*S, hidden)
    const Tensor input_matrix =
        framework::ReshapeToMatrix(*input, 2 /*x_num_col_dims */);
    // (hidden, 3 * all_head_size)
    const Tensor w_matrix =
        framework::ReshapeToMatrix(*w, 1 /*y_num_col_dims*/);

    Tensor temp_out_tensor;
    auto temp_out_dims =
        phi::make_ddim({batch, seq_len, 3, head_number, head_size});
    temp_out_tensor.Resize(
        {batch * seq_len, phi::product(temp_out_dims) / (batch * seq_len)});
    auto *temp_out_data = temp_out_tensor.mutable_data<T>(context.GetPlace());

    // (B * S, hidden) * (hidden, 3 * N * H) -> (B * S * 3 * N * H)
    auto blas = phi::funcs::GetBlas<phi::GPUContext, T>(device_ctx);
    blas.MatMul(input_matrix, w_matrix, &temp_out_tensor);

    // temp_out_tensor.Resize(temp_out_dims);

    Tensor multihead_temp_tensor;
    // B * head_number * S * S * 1 + B * S * 3 * N * H
    int scratch_size = batch * head_number * seq_len * seq_len * 1;
    multihead_temp_tensor.Resize({scratch_size + temp_out_tensor.numel()});
    auto *multihead_temp_data =
        multihead_temp_tensor.mutable_data<T>(context.GetPlace());
    auto *qkptr = multihead_temp_data;
    auto *tptr = multihead_temp_data + scratch_size;

    // Do the transpose with bias.
    // BxSx3xNxH => tptr: 3xBxNxSxH.
    TransQKVWithBias(batch,
                     seq_len,
                     head_size,
                     head_number,
                     temp_out_data,
                     bias_d,
                     tptr,
                     stream);
    if (std::is_same<T, platform::float16>::value) {
      math::MultiHeadGPUComputeFunctor<half> multihead_compute_func;
      multihead_compute_func(device_ctx,
                             batch,
                             seq_len,
                             head_number,
                             head_size,
                             reinterpret_cast<half *>(qkptr),
                             reinterpret_cast<const half *>(bias_qk_d),
                             reinterpret_cast<half *>(tptr),
                             __float2half(static_cast<float>(scale)),
                             __float2half(0.0));
    } else {
      math::MultiHeadGPUComputeFunctor<T> multihead_compute_func;
      multihead_compute_func(device_ctx,
                             batch,
                             seq_len,
                             head_number,
                             head_size,
                             qkptr,
                             bias_qk_d,
                             tptr,
                             scale,
                             T(0.0));
    }

    int grid = batch * head_number * seq_len;
    int block = head_size;
    transpose<T><<<grid, block, 0, stream>>>(
        tptr, output_d, batch, seq_len, head_number, head_size);
  }
};

}  // namespace operators
}  // namespace paddle

namespace ops = paddle::operators;
#if defined(PADDLE_WITH_CUDA) && CUDA_VERSION >= 10000
REGISTER_OP_CUDA_KERNEL(
    multihead_matmul,
    ops::MultiHeadMatMulV2Kernel<phi::GPUContext, paddle::platform::float16>,
    ops::MultiHeadMatMulV2Kernel<phi::GPUContext, float>);
#else
REGISTER_OP_CUDA_KERNEL(multihead_matmul,
                        ops::MultiHeadMatMulV2Kernel<phi::GPUContext, float>);
#endif