mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-11 17:00:18 +00:00
7f21662089
* Implement layernorm kernel and deviceOp * verify gpu kernel with host code * 1. Separate gamma aand beta from affine 2. Check if argument is valid * clean * Sync the naming * Support sweep once mode if we can put k dimension data inside one block * [What] Get length from upper length. [Why] if we get length directly, we may get length after padding. * We only use one block in K dimension. Hence, we can simplify the indexing of global R/W. * Use 1d descriptor for gamma and beta * Add accElementwiseOp * Extract layernorm host code * Support different YVectorDim in GridwiseLayernorm * Rename XSrcVectorDim to XYSrcVectorDim. Because we use same parameter in deviceOp * Gamma and beta can share the VGPR. * Add test for fp32 and fp16 * Fix bug of concurrency and add test case which may fail orignally * Propagate NaN for layernorm Co-authored-by: Chao Liu <chao.liu2@amd.com>
30 lines
1.9 KiB
C++
30 lines
1.9 KiB
C++
// SPDX-License-Identifier: MIT
|
|
// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
|
|
|
|
#include "gtest/gtest.h"
|
|
#include "test_layernorm_util.hpp"
|
|
|
|
template <ck::index_t N>
|
|
using I = ck::Number<N>;
|
|
|
|
template <typename Tuple>
|
|
class TestLayernormFP16 : public ck::TestLayernorm<Tuple>
|
|
{
|
|
};
|
|
|
|
// clang-format off
|
|
using KernelTypes = ::testing::Types<
|
|
// XDataType, GammaDataType, BetaDataType, AccDataType, YDataType, Rank, NumReduceDim, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, XYSrcVectorDim, XSrcVectorSize, , GammaSrcVectorSize, BetaSrcVectorSize, YDstVectorSize>
|
|
std::tuple<ck::half_t, ck::half_t, ck::half_t, float, ck::half_t, I<2>, I<1>, I<256>, I<8>, I<32>, I<1>, I<8>, I<1>, I<8>, I<8>, I<8>, I<8>>,
|
|
std::tuple<ck::half_t, ck::half_t, ck::half_t, float, ck::half_t, I<2>, I<1>, I<256>, I<8>, I<32>, I<2>, I<8>, I<1>, I<8>, I<8>, I<8>, I<8>>,
|
|
std::tuple<ck::half_t, ck::half_t, ck::half_t, float, ck::half_t, I<2>, I<1>, I<256>, I<4>, I<64>, I<1>, I<8>, I<1>, I<8>, I<8>, I<8>, I<8>>,
|
|
std::tuple<ck::half_t, ck::half_t, ck::half_t, float, ck::half_t, I<2>, I<1>, I<256>, I<4>, I<64>, I<2>, I<8>, I<1>, I<8>, I<8>, I<8>, I<8>>,
|
|
std::tuple<ck::half_t, ck::half_t, ck::half_t, float, ck::half_t, I<2>, I<1>, I<256>, I<2>, I<128>, I<1>, I<8>, I<1>, I<8>, I<8>, I<8>, I<8>>,
|
|
std::tuple<ck::half_t, ck::half_t, ck::half_t, float, ck::half_t, I<2>, I<1>, I<256>, I<2>, I<128>, I<2>, I<8>, I<1>, I<8>, I<8>, I<8>, I<8>>,
|
|
std::tuple<ck::half_t, ck::half_t, ck::half_t, float, ck::half_t, I<2>, I<1>, I<256>, I<1>, I<256>, I<1>, I<8>, I<1>, I<8>, I<8>, I<8>, I<8>>,
|
|
std::tuple<ck::half_t, ck::half_t, ck::half_t, float, ck::half_t, I<2>, I<1>, I<256>, I<1>, I<256>, I<2>, I<8>, I<1>, I<8>, I<8>, I<8>, I<8>>
|
|
>;
|
|
// clang-format on
|
|
TYPED_TEST_SUITE(TestLayernormFP16, KernelTypes);
|
|
TYPED_TEST(TestLayernormFP16, Test_FP16) { this->Run(); }
|