minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

[CPU] remove RECORD_FUNCTION (#23528)

Browse Source
This commit is contained in:
Ma Mingfei
2026-04-24 09:18:29 +08:00
committed by GitHub
parent 87e50f20f6
commit 23e4d381f0
21 changed files with 0 additions and 106 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sgl-kernel/csrc/cpu/norm.cpp
-17
View File
@@ -538,8 +538,6 @@ void fused_add_layernorm_kernel_impl(
// input : {batch_size, hidden_size}
at::Tensor l2norm_cpu(at::Tensor& input, double eps) {
RECORD_FUNCTION("sgl-kernel::l2norm_cpu", std::vector<c10::IValue>({input}));
CHECK_INPUT(input);
CHECK_DIM(2, input);
int64_t batch_size = input.size(0);
@@ -565,8 +563,6 @@ at::Tensor l2norm_cpu(at::Tensor& input, double eps) {
// input : {batch_size, hidden_size} or {batch_size, seq_len, hidden_size}
// weight: {hidden_size}
at::Tensor rmsnorm_cpu(at::Tensor& input, at::Tensor& weight, double eps) {
RECORD_FUNCTION("sgl-kernel::rmsnorm_cpu", std::vector<c10::IValue>({input, weight}));
CHECK_LAST_DIM_CONTIGUOUS_INPUT(input);
CHECK_INPUT(weight);
int64_t inp_dim{input.dim()};
@@ -613,8 +609,6 @@ at::Tensor rmsnorm_cpu(at::Tensor& input, at::Tensor& weight, double eps) {
// bias : {hidden_size}
at::Tensor
layernorm_cpu(const at::Tensor& input, const at::Tensor& weight, const std::optional<at::Tensor>& bias, double eps) {
RECORD_FUNCTION("sgl-kernel::layernorm_cpu", std::vector<c10::IValue>({input, weight}));
CHECK_LAST_DIM_CONTIGUOUS_INPUT(input);
CHECK_INPUT(weight);
int64_t inp_dim{input.dim()};
@@ -657,8 +651,6 @@ layernorm_cpu(const at::Tensor& input, const at::Tensor& weight, const std::opti
}
at::Tensor gemma_rmsnorm_cpu(at::Tensor& input, at::Tensor& weight, double eps) {
RECORD_FUNCTION("sgl-kernel::gemma_rmsnorm_cpu", std::vector<c10::IValue>({input, weight}));
CHECK_LAST_DIM_CONTIGUOUS_INPUT(input);
CHECK_INPUT(weight);
CHECK_DIM(2, input);
@@ -694,8 +686,6 @@ at::Tensor gemma_rmsnorm_cpu(at::Tensor& input, at::Tensor& weight, double eps)
// input : {batch_size, hidden_size} or {batch_size, num_head, seq_len, head_dim}
// weight: {hidden_size}
at::Tensor gemma3_rmsnorm_cpu(at::Tensor& input, at::Tensor& weight, double eps) {
RECORD_FUNCTION("sgl-kernel::gemma3_rmsnorm_cpu", std::vector<c10::IValue>({input, weight}));
CHECK_LAST_DIM_CONTIGUOUS_INPUT(input);
CHECK_INPUT(weight);
TORCH_CHECK(
@@ -761,8 +751,6 @@ at::Tensor gemma3_rmsnorm_cpu(at::Tensor& input, at::Tensor& weight, double eps)
// input : {batch_size, hidden_size} or {batch_size, seq_len, hidden_size}
// weight: {hidden_size}
at::Tensor gemma4_rmsnorm_cpu(at::Tensor& input, at::Tensor& weight, double eps, double scale_shift, bool with_scale) {
RECORD_FUNCTION("sgl-kernel::gemma4_rmsnorm_cpu", std::vector<c10::IValue>({input, weight}));
CHECK_LAST_DIM_CONTIGUOUS_INPUT(input);
CHECK_INPUT(weight);
int64_t inp_dim{input.dim()};
@@ -826,8 +814,6 @@ at::Tensor gemma4_rmsnorm_cpu(at::Tensor& input, at::Tensor& weight, double eps,
// weight: {hidden_size}
// gate: {batch_size, hidden_size}
at::Tensor fused_rmsnorm_gated_cpu(at::Tensor& input, at::Tensor& weight, at::Tensor& gate, double eps) {
RECORD_FUNCTION("sgl-kernel::fused_rmsnorm_gated_cpu", std::vector<c10::IValue>({input, weight, gate}));
CHECK_LAST_DIM_CONTIGUOUS_INPUT(input);
CHECK_INPUT(weight);
CHECK_INPUT(gate);
@@ -860,7 +846,6 @@ at::Tensor fused_rmsnorm_gated_cpu(at::Tensor& input, at::Tensor& weight, at::Te
// residual: {batch_size, hidden_size} or {batch_size, seq_len, hidden_size}
// weight : {hidden_size}
void fused_add_rmsnorm_cpu(at::Tensor& input, at::Tensor& residual, at::Tensor& weight, double eps) {
RECORD_FUNCTION("sgl-kernel::fused_add_rmsnorm_cpu", std::vector<c10::IValue>({input, residual, weight}));
CHECK_LAST_DIM_CONTIGUOUS_INPUT(input);
CHECK_INPUT(residual);
CHECK_INPUT(weight);
@@ -909,7 +894,6 @@ void fused_add_rmsnorm_cpu(at::Tensor& input, at::Tensor& residual, at::Tensor&
// residual: {batch_size, hidden_size}
// weight : {hidden_size}
void gemma_fused_add_rmsnorm_cpu(at::Tensor& input, at::Tensor& residual, at::Tensor& weight, double eps) {
RECORD_FUNCTION("sgl-kernel::gemma_fused_add_rmsnorm_cpu", std::vector<c10::IValue>({input, residual, weight}));
CHECK_LAST_DIM_CONTIGUOUS_INPUT(input);
CHECK_INPUT(residual);
CHECK_INPUT(weight);
@@ -957,7 +941,6 @@ at::Tensor fused_add_layernorm_cpu(
const at::Tensor& weight,
const std::optional<at::Tensor>& bias,
double eps) {
RECORD_FUNCTION("sgl-kernel::fused_add_layernorm_cpu", std::vector<c10::IValue>({input, residual, weight}));
CHECK_LAST_DIM_CONTIGUOUS_INPUT(input);
CHECK_INPUT(residual);
CHECK_INPUT(weight);