Apply suggestions from code review

Accept suggestions from liji.

Co-authored-by: Jin Li <[email protected]>
Signed-off-by: dongxuy04 <[email protected]>

Author: dongxuy04

cpp/tensorrt_llm/thop/moeCommOp.cpp

@@ -242,7 +242,7 @@ TORCH_LIBRARY_IMPL(trtllm, CUDA, m)
 
 TORCH_LIBRARY_FRAGMENT(trtllm, m)
 {
-    m.def("moe_initialize_workspace(Tensor all_workspaces, int ep_rank, int ep_size) -> ()");
+    m.def("moe_initialize_workspace(Tensor(a!) all_workspaces, int ep_rank, int ep_size) -> ()");
 }
 
 TORCH_LIBRARY_IMPL(trtllm, CUDA, m)
@@ -286,7 +286,7 @@ TORCH_LIBRARY_IMPL(trtllm, CUDA, m)
 TORCH_LIBRARY_FRAGMENT(trtllm, m)
 {
     m.def(
-        "memset_expert_ids(Tensor experts_ids, Tensor recv_rank_count_cumsum, int max_token_count_per_rank, int top_k, "
+        "memset_expert_ids(Tensor(a!) experts_ids, Tensor recv_rank_count_cumsum, int max_token_count_per_rank, int top_k, "
         "int slot_count, int ep_size) -> ()");
 }