github
/
ColossalAI
mirror of https://github.com/hpcaitech/ColossalAI
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

[NFC] polish colossalai/kernel/cuda_native/csrc/scaled_upper_triang_masked_softmax.cpp code style (#959)

pull/997/head
superhao1995 2022-05-15 09:01:34 +08:00 committed by binmakeswell
parent 442a2975ab
commit 48c4a180c7
1 changed files with 20 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

45
colossalai/kernel/cuda_native/csrc/scaled_upper_triang_masked_softmax.cpp
View File

@ -3,57 +3,52 @@
#include <cuda_fp16.h>
#include <torch/extension.h>
#include <vector>
namespace multihead_attn {
namespace fused_softmax {
namespace scaled_upper_triang_masked_softmax {
torch::Tensor fwd_cuda(
torch::Tensor const& input,
float scale_factor);
torch::Tensor fwd_cuda(torch::Tensor const& input, float scale_factor);
torch::Tensor bwd_cuda(
torch::Tensor const& output_grads,
torch::Tensor const& softmax_results,
float scale_factor);
torch::Tensor bwd_cuda(torch::Tensor const& output_grads,
torch::Tensor const& softmax_results,
float scale_factor);
torch::Tensor fwd(torch::Tensor const& input, float scale_factor) {
AT_ASSERTM(input.dim() == 3, "expected 3D tensor");
AT_ASSERTM((input.scalar_type() == at::ScalarType::Half) ||
(input.scalar_type() == at::ScalarType::BFloat16),
"Only fp16 and bf16 are supported");
(input.scalar_type() == at::ScalarType::BFloat16),
"Only fp16 and bf16 are supported");
return fwd_cuda(input, scale_factor);
}
torch::Tensor bwd(
torch::Tensor const& output_grads,
torch::Tensor const& softmax_results,
float scale_factor) {
torch::Tensor bwd(torch::Tensor const& output_grads,
torch::Tensor const& softmax_results, float scale_factor) {
AT_ASSERTM(output_grads.dim() == 3, "expected 3D tensor");
AT_ASSERTM(softmax_results.dim() == 3, "expected 3D tensor");
AT_ASSERTM((output_grads.scalar_type() == at::ScalarType::Half) ||
(output_grads.scalar_type() == at::ScalarType::BFloat16),
"Only fp16 and bf16 are supported");
(output_grads.scalar_type() == at::ScalarType::BFloat16),
"Only fp16 and bf16 are supported");
AT_ASSERTM((softmax_results.scalar_type() == at::ScalarType::Half) ||
(softmax_results.scalar_type() == at::ScalarType::BFloat16),
"Only fp16 and bf16 are supported");
(softmax_results.scalar_type() == at::ScalarType::BFloat16),
"Only fp16 and bf16 are supported");
return bwd_cuda(output_grads, softmax_results, scale_factor);
}
} // end namespace scaled_upper_triang_masked_softmax
} // end namespace fused_softmax
} // end namespace multihead_attn
} // end namespace scaled_upper_triang_masked_softmax
} // end namespace fused_softmax
} // end namespace multihead_attn
PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
m.def("forward",
m.def("forward",
&multihead_attn::fused_softmax::scaled_upper_triang_masked_softmax::fwd,
"Self Multihead Attention scaled, time masked softmax -- Forward.");
m.def("backward",
"Self Multihead Attention scaled, time masked softmax -- Forward.");
m.def("backward",
&multihead_attn::fused_softmax::scaled_upper_triang_masked_softmax::bwd,
"Self Multihead Attention scaled, time masked softmax -- Backward.");
"Self Multihead Attention scaled, time masked softmax -- Backward.");
}