fix lint problems

Author: luanyundu

csrc/deepep/deep_ep.cpp

@@ -77,7 +77,7 @@ Buffer::get_dispatch_layout(const torch::Tensor &topk_idx, int num_experts, std:
 
     const int num_tokens = new_topk_idx.size(0);
     const int num_topk = new_topk_idx.size(1);
-    const int local_ranksize = A2_LOCAL_RANK_SIZE;
+    const int local_ranksize = LOCAL_RANK_SIZE;
     auto server_num = num_ranks / local_ranksize;
 
     auto device = new_topk_idx.device();
@@ -87,15 +87,24 @@ Buffer::get_dispatch_layout(const torch::Tensor &topk_idx, int num_experts, std:
     const int notify_send_data_size =
         num_experts * EXPERT_DATA_SIZE + server_num + MAX_BATCH_SIZE * (1 + 2 * server_num + num_topk);
     /*
-    The output parameters are ordered as follows:
-    1. the number of the tokens that every expert received from this NPU.                size:[numExpert]
-    2. The number of tokens received by each server from this NPU (deduplicated).        size:[serverNum]
-    3. The number of tokens sent from this NPU to each server (without deduplication).   size:[MAX_BS, serverNum]
-    4. The number of servers each token is sent to by this NPU.                          size:[MAX_BS]
-    5. The order in which each token of this NPU is sent to various servers.             size:[MAX_BS, serverNum]
-    6. The order in which each token is sent to the expert.                              size:[MAX_BS, numTopk]
-    7. The server offset of tokens received by each expert from this NPU.                size:[numExpert, MAX_BS]
-    8. The origin offset of the token received by each expert on the original NPU.       size:[numExpert, MAX_BS]
+    The notify send data is constructed by 8 parameters and
+    the parameters are ordered as follows:
+    1. the number of the tokens that every expert received from this NPU.
+       size:[numExpert]
+    2. The number of tokens received by each server from this NPU (deduplicated).
+       size:[serverNum]
+    3. The number of tokens sent from this NPU to each server (without deduplication).
+       size:[MAX_BS, serverNum]
+    4. The number of servers each token is sent to by this NPU.
+       size:[MAX_BS]
+    5. The order in which each token of this NPU is sent to various servers.
+       size:[MAX_BS, serverNum]
+    6. The order in which each token is sent to the expert.
+       size:[MAX_BS, numTopk]
+    7. The server offset of tokens received by each expert from this NPU.
+       size:[numExpert, MAX_BS]
+    8. The origin offset of the token received by each expert on the original NPU.
+       size:[numExpert, MAX_BS]
     */
     auto notify_send_data = at::zeros({notify_send_data_size}, at::dtype(at::kInt).device(device));
     notify_send_data