fix a split index bug in launch allreduce

mindspore/ccsrc/backend/session/session_basic.cc

@@ -2313,6 +2313,23 @@ std::vector<uint32_t> GenerateBucketSizeList(const KernelGraphPtr &graph, const
   return bucket_size_list;
 }
 
+void PreProcessOnSplitIndex(const KernelGraphPtr &graph, vector<uint32_t> *split_index) {
+  if (split_index->empty()) {
+    return;
+  }
+  // calculate split index num
+  auto split_index_len = split_index->size();
+  uint32_t split_index_num = (*split_index)[split_index_len - 1];
+  // obtain graph output tensor num
+  auto grads_count = GetBpropGraphGradsCount(graph);
+  if (split_index_num == 0 || split_index_num >= grads_count) {
+    MS_LOG(EXCEPTION) << "invalid AllReduce split index " << split_index_num << " and grads count " << grads_count;
+  }
+  if (split_index_num < grads_count - 1) {
+    split_index->push_back(grads_count - 1);
+  }
+}
+
 void SessionBasic::InitAllBucket(const KernelGraphPtr &graph) {
   MS_EXCEPTION_IF_NULL(graph);
   MS_LOG(INFO) << "Init Bucket start, graph_id:" << graph->graph_id();
@@ -2325,6 +2342,7 @@ void SessionBasic::InitAllBucket(const KernelGraphPtr &graph) {
   std::vector<std::shared_ptr<device::Bucket>> bucket_list;
   // Create bucket for every split allreduce ops
   auto split_index = GetAllReduceSplitIndex();
+  PreProcessOnSplitIndex(graph, &split_index);
   auto bucket_size_list = GenerateBucketSizeList(graph, split_index);
   uint32_t bucket_id = 0;
   for (auto bucket_size : bucket_size_list) {