0dc3ba60b3
* Support half precision sigmoid activation * introduce a vectorized variant using fast_tanh * refactored sigmoid using the new interface * refactored gelu * add silu activation * add hardswish * remove sigmoid for now * add description to silu and hardswish, and other doc update * Do not ignore Round * use constant N * Set isHeavy = true in sigmoid and silu epilogue
65 lines
3.4 KiB
C++
65 lines
3.4 KiB
C++
/***************************************************************************************************
|
|
* Copyright (c) 2017-2021, NVIDIA CORPORATION. All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without modification, are permitted
|
|
* provided that the following conditions are met:
|
|
* * Redistributions of source code must retain the above copyright notice, this list of
|
|
* conditions and the following disclaimer.
|
|
* * Redistributions in binary form must reproduce the above copyright notice, this list of
|
|
* conditions and the following disclaimer in the documentation and/or other materials
|
|
* provided with the distribution.
|
|
* * Neither the name of the NVIDIA CORPORATION nor the names of its contributors may be used
|
|
* to endorse or promote products derived from this software without specific prior written
|
|
* permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR
|
|
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
|
|
* FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
|
|
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
|
|
* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
|
|
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
|
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
*
|
|
**************************************************************************************************/
|
|
/*! \file
|
|
\brief Functor performing linear combination with GELU operations used by epilogues.
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include "cutlass/cutlass.h"
|
|
#include "cutlass/epilogue/thread/activation.h"
|
|
#include "cutlass/epilogue/thread/linear_combination_generic.h"
|
|
|
|
/////////////////////////////////////////////////////////////////////////////////////////////////
|
|
|
|
namespace cutlass {
|
|
namespace epilogue {
|
|
namespace thread {
|
|
|
|
/////////////////////////////////////////////////////////////////////////////////////////////////
|
|
|
|
/// Applies a linear combination operator followed by the GELU activation to an array of elements.
|
|
///
|
|
/// D = gelu(alpha * accumulator + beta * source + uniform)
|
|
///
|
|
template <
|
|
typename ElementOutput_, ///< Data type used to load and store tensors
|
|
int Count, ///< Number of elements computed per operation
|
|
///< Usually it is 128/sizeof_bits<ElementOutput_>,
|
|
///< but we use 64 or 32 sometimes when there are not enough data to store
|
|
typename ElementAccumulator_ = ElementOutput_, ///< Accumulator data type
|
|
typename ElementCompute_ = ElementOutput_, ///< Data type used to compute linear combination
|
|
FloatRoundStyle Round = FloatRoundStyle::round_to_nearest
|
|
>
|
|
using LinearCombinationGELU = LinearCombinationGeneric<GELU, ElementOutput_, Count, ElementAccumulator_,
|
|
ElementCompute_, Round, true>;
|
|
|
|
|
|
/////////////////////////////////////////////////////////////////////////////////////////////////
|
|
|
|
} // namespace thread
|
|
} // namespace epilogue
|
|
} // namespace cutlass
|