ColossalAI/colossalai/kernel/cuda_native/csrc/colossal_C_frontend.cpp

49 lines
2.5 KiB
C++
Raw Normal View History

2022-03-31 07:13:01 +00:00
// modified from
// https://github.com/NVIDIA/apex/blob/master/csrc/multi_tensor_adam.cu
2021-10-28 16:21:23 +00:00
#include <torch/extension.h>
2022-03-31 07:13:01 +00:00
void multi_tensor_scale_cuda(int chunk_size, at::Tensor noop_flag,
std::vector<std::vector<at::Tensor>> tensor_lists,
float scale);
2021-10-28 16:21:23 +00:00
2022-03-31 07:13:01 +00:00
void multi_tensor_sgd_cuda(int chunk_size, at::Tensor noop_flag,
std::vector<std::vector<at::Tensor>> tensor_lists,
float wd, float momentum, float dampening, float lr,
bool nesterov, bool first_run,
bool wd_after_momentum, float scale);
2021-10-28 16:21:23 +00:00
2022-03-31 07:13:01 +00:00
void multi_tensor_adam_cuda(int chunk_size, at::Tensor noop_flag,
std::vector<std::vector<at::Tensor>> tensor_lists,
const float lr, const float beta1,
const float beta2, const float epsilon,
const int step, const int mode,
const int bias_correction,
const float weight_decay);
2021-10-28 16:21:23 +00:00
2022-03-31 07:13:01 +00:00
void multi_tensor_lamb_cuda(int chunk_size, at::Tensor noop_flag,
std::vector<std::vector<at::Tensor>> tensor_lists,
const float lr, const float beta1,
const float beta2, const float epsilon,
const int step, const int bias_correction,
const float weight_decay, const int grad_averaging,
const int mode, at::Tensor global_grad_norm,
const float max_grad_norm,
at::optional<bool> use_nvlamb_python);
2021-10-28 16:21:23 +00:00
std::tuple<at::Tensor, at::Tensor> multi_tensor_l2norm_cuda(
int chunk_size, at::Tensor noop_flag,
std::vector<std::vector<at::Tensor>> tensor_lists,
at::optional<bool> per_tensor_python);
2021-10-28 16:21:23 +00:00
2022-03-31 07:13:01 +00:00
PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
m.def("multi_tensor_scale", &multi_tensor_scale_cuda,
2021-10-28 16:21:23 +00:00
"Fused overflow check + scale for a list of contiguous tensors");
2022-03-31 07:13:01 +00:00
m.def("multi_tensor_sgd", &multi_tensor_sgd_cuda,
"Fused SGD optimizer for list of contiguous tensors");
m.def("multi_tensor_adam", &multi_tensor_adam_cuda,
"Compute and apply gradient update to parameters for Adam optimizer");
m.def("multi_tensor_lamb", &multi_tensor_lamb_cuda,
"Computes and apply update for LAMB optimizer");
m.def("multi_tensor_l2norm", &multi_tensor_l2norm_cuda,
"Computes L2 norm for a list of contiguous tensors");
2021-10-28 16:21:23 +00:00
}