mirror of
https://github.com/hpcaitech/ColossalAI.git
synced 2025-05-05 06:58:09 +00:00
279300dc5f
* refactor compilation mechanism and unified multi hw * fix file path bug * add init.py to make pybind a module to avoid relative path error caused by softlink * delete duplicated micros * fix micros bug in gcc
78 lines
2.6 KiB
Plaintext
78 lines
2.6 KiB
Plaintext
#include <ATen/cuda/CUDAContext.h>
|
||
#include <torch/extension.h>
|
||
#include <stdio.h>
|
||
|
||
#include "common/micros.h"
|
||
#include "common/mp_type_traits.h"
|
||
|
||
using colossalAI::common::MPTypeTrait;
|
||
|
||
template<typename T>
|
||
__device__ __forceinline__ T silu_kernel(const T& x) {
|
||
// x * sigmoid(x)
|
||
using MT = typename MPTypeTrait<T>::Type;
|
||
return static_cast<T>((static_cast<MT>(x)) / (static_cast<MT>(1.0f) + expf(static_cast<MT>(-x))));
|
||
}
|
||
|
||
template<typename scalar_t, scalar_t (*ACT_FN)(const scalar_t&)>
|
||
__global__ void act_and_mul_kernel(
|
||
const scalar_t* __restrict__ ins_data,
|
||
scalar_t* __restrict__ outs_data,
|
||
const int64_t numel) {
|
||
using MT = typename MPTypeTrait<scalar_t>::Type;
|
||
|
||
int64_t idx = static_cast<int64_t>(threadIdx.x) + static_cast<int64_t>(blockIdx.x) * static_cast<int64_t>(blockDim.x);
|
||
const int64_t grid_size = blockDim.x * gridDim.x;
|
||
if(idx > numel) {
|
||
return;
|
||
}
|
||
|
||
for(int64_t i = idx; i < numel; i += grid_size) {
|
||
scalar_t x = ins_data[i];
|
||
scalar_t y = ins_data[i+numel];
|
||
outs_data[i] = static_cast<scalar_t>(static_cast<MT>(ACT_FN(x)) * static_cast<MT>(y));
|
||
}
|
||
}
|
||
|
||
// Note(LiuYang):This func is designed for calculation mode like
|
||
// silu(x[:half_1stdim]) * (x[half_1stdim:])
|
||
torch::Tensor silu_and_mul(const torch::Tensor& ins)
|
||
{
|
||
// Note(LiuYang): According to torch doc, vec() may cost a lot, but I did't find a better api
|
||
// to manipulate ins_shape which is IntArrayRef
|
||
auto ins_shape = ins.sizes().vec();
|
||
|
||
ins_shape[0] = ins_shape[0]/2;
|
||
if (ins_shape[0] == 1) {
|
||
ins_shape.erase(ins_shape.begin());
|
||
}
|
||
auto outs = torch::zeros(ins_shape,ins.options());
|
||
|
||
const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
|
||
|
||
// Note(Liuyang): numel of ins must be divisible by 2
|
||
int64_t numel = ((torch::numel(ins)) >> 1);
|
||
|
||
// Note(LiuYang): For better performance for special case of which input is [2, 64, 11008], now
|
||
// I comment this part code，because it also cost a little time to calculate a better config
|
||
// colossalAI::cuda::utils::NVGPUDevInfo dev_info(0);
|
||
// auto config = colossalAI::cuda::utils::GetGPULaunchConfig1D(dev_info,numel,1);
|
||
// dim3 grid = config.grid;
|
||
// dim3 block = config.block;
|
||
|
||
dim3 grid((numel+255)/256);
|
||
dim3 block(256);
|
||
|
||
DISPATCH_FLOAT_HALF_AND_BFLOAT(
|
||
ins.scalar_type(),
|
||
"silu_and_mul",
|
||
act_and_mul_kernel<scalar_t,silu_kernel<scalar_t>><<<grid, block, 0, stream>>>(
|
||
ins.data_ptr<scalar_t>(),
|
||
outs.data_ptr<scalar_t>(),
|
||
numel
|
||
);)
|
||
|
||
AT_CUDA_CHECK(cudaGetLastError());
|
||
return outs;
|
||
}
|