diff --git a/sgl-kernel/csrc/common_extension_rocm.cc b/sgl-kernel/csrc/common_extension_rocm.cc
index 54c520641..2868153d5 100644
--- a/sgl-kernel/csrc/common_extension_rocm.cc
+++ b/sgl-kernel/csrc/common_extension_rocm.cc
@@ -165,6 +165,16 @@ TORCH_LIBRARY_EXPAND(sgl_kernel, m) {
       "transfer_kv_all_layer_direct_lf_pf(Tensor[] src_ptrs, Tensor[] dst_ptrs, Tensor src_indices, "
       "Tensor dst_indices, int page_size) ->() ");
   m.impl("transfer_kv_all_layer_direct_lf_pf", torch::kCUDA, &transfer_kv_all_layer_direct_lf_pf);
+  m.def(
+      "transfer_kv_all_layer_lf_ph(Tensor src_k_layers, Tensor dst_k, Tensor src_v_layers, Tensor dst_v, "
+      "Tensor src_indices, Tensor dst_indices, int item_size, int dst_layout_dim, int num_layers, int page_size, int "
+      "head_num, int block_quota, int num_warps_per_block) -> ()");
+  m.impl("transfer_kv_all_layer_lf_ph", torch::kCUDA, &transfer_kv_all_layer_lf_ph);
+  m.def(
+      "transfer_kv_per_layer_ph_lf(Tensor src_k, Tensor dst_k, Tensor src_v, Tensor dst_v, Tensor src_indices, Tensor "
+      "dst_indices, int layer_id, int item_size, int src_layout_dim, int page_size, int head_num, int block_quota, int "
+      "num_warps_per_block) -> ()");
+  m.impl("transfer_kv_per_layer_ph_lf", torch::kCUDA, &transfer_kv_per_layer_ph_lf);
 
   /*
    * From csrc/grammar