!8895 [MSLITE]improve train code

From: @guohonhzilonghw
Reviewed-by: @ddwsky,@zhanghaibo5
Signed-off-by: @zhanghaibo5

mindspore/lite/src/train/loss_kernel.h

@@ -22,9 +22,9 @@ namespace mindspore::kernel {
 class LossKernel : public LiteKernel {
  public:
   LossKernel() = default;
-  explicit LossKernel(OpParameter *parameter, const std::vector<lite::Tensor *> &inputs,
-                      const std::vector<lite::Tensor *> &outputs, const lite::InnerContext *ctx,
-                      const lite::PrimitiveC *primitive)
+  LossKernel(OpParameter *parameter, const std::vector<lite::Tensor *> &inputs,
+             const std::vector<lite::Tensor *> &outputs, const lite::InnerContext *ctx,
+             const lite::PrimitiveC *primitive)
       : LiteKernel(parameter, inputs, outputs, ctx, primitive) {}
   ~LossKernel() = default;
 };

mindspore/lite/src/train/train_model.cc

@@ -49,12 +49,14 @@ TrainModel *TrainModel::Import(const char *model_buf, size_t size) {
   const void *meta_graph = GetMetaGraphByVerison(model->buf, schema_version);
   if (meta_graph == nullptr) {
     MS_LOG(ERROR) << "meta_graph is nullptr!";
+    free(model->buf);
     delete (model);
     return nullptr;
   }
 
   int status = GenerateModelByVersion(meta_graph, model, schema_version);
   if (status != RET_OK) {
+    free(model->buf);
     delete (model);
     MS_LOG(ERROR) << "fail to generate model";
     return nullptr;
@@ -73,17 +75,16 @@ char *TrainModel::ExportBuf(char *buffer, size_t *len) const {
     MS_LOG(ERROR) << "Model::Export is only available for Train Session";
     return nullptr;
   }
-
   if (*len < buf_size_ && buffer != nullptr) {
     MS_LOG(ERROR) << "Buffer is too small, Export Failed";
     return nullptr;
   }
   if (buffer == nullptr) {
     buffer = reinterpret_cast<char *>(malloc(buf_size_));
-  }
-  if (buffer == nullptr) {
-    MS_LOG(ERROR) << "allocated model buf fail!";
-    return nullptr;
+    if (buffer == nullptr) {
+      MS_LOG(ERROR) << "allocated model buf fail!";
+      return nullptr;
+    }
   }
 
   memcpy(buffer, buf, buf_size_);

mindspore/lite/src/train/train_session.cc

@@ -92,13 +92,22 @@ void TrainSession::AllocWorkSpace() {
 int TrainSession::CompileGraph(lite::Model *model) { return lite::RET_ERROR; }
 
 int TrainSession::CompileTrainGraph(mindspore::lite::TrainModel *model) {
+  if (model == nullptr) {
+    MS_LOG(ERROR) << "model is null";
+    return RET_ERROR;
+  }
   model_ = model;
-
   auto restore = ReplaceOps();
   auto ret = lite::LiteSession::CompileGraph(model);
+  if (ret != RET_OK) {
+    MS_LOG(ERROR) << "Compile train graph failed";
+    return RET_ERROR;
+  }
   orig_output_map_ = output_node_map_;
   orig_output_tensor_map_ = output_tensor_map_;
-  for (auto inTensor : inputs_) inTensor->MutableData();
+  for (auto inTensor : inputs_) {
+    inTensor->MutableData();
+  }
   RestoreOps(restore);
   AllocWorkSpace();
   MarkOptimizedKernels();
@@ -152,7 +161,7 @@ int TrainSession::RunGraph(const KernelCallBack &before, const KernelCallBack &a
 int TrainSession::SaveToFile(const std::string &filename) const {
   size_t fb_size = 0;
   auto *buf = reinterpret_cast<char *>(ExportToBuf(nullptr, &fb_size));
-  if (buf == NULL) {
+  if (buf == nullptr) {
     MS_LOG(ERROR) << "Could not Export Trained model";
     return lite::RET_NULL_PTR;
   }
@@ -212,7 +221,7 @@ int TrainSession::Train() {
 }
 
 void TrainSession::UpdateOutputMapByLossKernel(const kernel::LiteKernel *kernel) {
-  if (IsLossKernel(kernel)) {
+  if (kernel != nullptr && IsLossKernel(kernel)) {
     auto *ms_tensor = kernel->out_tensors().at(0);
     if (ms_tensor != nullptr) {
       (void)ms_tensor->MutableData();
@@ -226,7 +235,7 @@ void TrainSession::UpdateOutputMapByLossKernel(const kernel::LiteKernel *kernel)
 }
 
 void TrainSession::UpdateOutputMapByInKernel(const kernel::LiteKernel *kernel) {
-  if (IsLossKernel(kernel)) {
+  if (kernel != nullptr && IsLossKernel(kernel)) {
     for (auto in_kernel : kernel->in_kernels()) {
       if (output_node_map_.find(in_kernel->name()) == output_node_map_.end()) {
         auto *ms_tensor = in_kernel->out_tensors().at(0);
@@ -304,9 +313,9 @@ void TrainSession::BuildInferenceKernelsMap() {
       }
     } else {
       auto sub_graph = reinterpret_cast<kernel::SubGraphKernel *>(kernel);
-      for (auto sb_kernel : sub_graph->nodes()) {
-        if (IsLossKernel(sb_kernel)) {  // For each loss in the system add backward tree
-          for (auto in_node : sb_kernel->in_kernels()) {
+      for (auto sub_kernel : sub_graph->nodes()) {
+        if (IsLossKernel(sub_kernel)) {  // For each loss in the system add backward tree
+          for (auto in_node : sub_kernel->in_kernels()) {
             BuildInferenceKernelsRecursive(in_node, &req_kernels);
           }
         }
@@ -357,9 +366,9 @@ void TrainSession::MarkOptimizedKernels() {
       }
     } else {
       auto sub_graph = reinterpret_cast<kernel::SubGraphKernel *>(kernel);
-      for (auto sb_kernel : sub_graph->nodes()) {
-        if (IsOptimizer(sb_kernel)) {
-          std::copy(sb_kernel->in_tensors().begin(), sb_kernel->in_tensors().end(), std::back_inserter(ot));
+      for (auto sub_kernel : sub_graph->nodes()) {
+        if (IsOptimizer(sub_kernel)) {
+          std::copy(sub_kernel->in_tensors().begin(), sub_kernel->in_tensors().end(), std::back_inserter(ot));
         }
       }
     }
@@ -376,11 +385,11 @@ void TrainSession::MarkOptimizedKernels() {
       }
     } else {
       auto sub_graph = reinterpret_cast<kernel::SubGraphKernel *>(kernel);
-      for (auto sb_kernel : sub_graph->nodes()) {
-        if (!IsOptimizer(sb_kernel)) {
-          for (auto it : sb_kernel->in_tensors()) {
+      for (auto sub_kernel : sub_graph->nodes()) {
+        if (!IsOptimizer(sub_kernel)) {
+          for (auto it : sub_kernel->in_tensors()) {
             if (std::find(ot.begin(), ot.end(), it) != ot.end()) {
-              sb_kernel->set_trainable(true);
+              sub_kernel->set_trainable(true);
               break;
             }
           }