apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[XPU] refactor of block_attn param 'pos_emb_type' (#5511)

Browse Source
This commit is contained in:
Lucas
2025-12-12 14:30:09 +08:00
committed by GitHub
parent 4eb55332f6
commit 888c4b992d
6 changed files with 25 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
custom_ops/xpu_ops/src/ops/pybind/pybind.cc
View File

@@ -85,8 +85,8 @@ std::vector<paddle::Tensor> BlockAttn(
const paddle::optional<paddle::Tensor>& smooth,
const paddle::optional<paddle::Tensor>& kv_signal_data_cpu,
const paddle::optional<paddle::Tensor>& cachekv_signal_thread_cpu,
const std::string& pos_emb_type = "NORMAL",
bool rope_3d = false);
const bool use_neox_rotary_style,
const bool rope_3d = false);
std::vector<paddle::Tensor> MoeLayer(
const paddle::Tensor& x,
@@ -616,7 +616,7 @@ PYBIND11_MODULE(fastdeploy_ops, m) {
py::arg("smooth"),
py::arg("kv_signal_data_cpu"),
py::arg("cachekv_signal_thread_cpu"),
py::arg("pos_emb_type") = "NORMAL",
py::arg("use_neox_rotary_style"),
py::arg("rope_3d") = false,
"block attention in XPU");