Caffe Python MemoryDataLayer Segmentation Fault

轉載請註明出處，樓燚(yì)航的blog，http://home.cnblogs.com/louyihang-loves-baiyan/

由於利用Pyhon來作數據的預處理比較方便，所以在data_layer選擇上，採用了MemoryDataLayer,能夠比較方便的直接用Python 根據set_input_array進行feed數據，而後再調用solver進行step就能夠了。說一下我碰到的問題，當時檢查了一下感受沒有哪裏出錯，可是報

Segmentation Fault(Core Abort)

感受好囧，最怕這個了。通常說段錯誤都是內存錯誤，好比數組越界，無效的指針，引用被釋放的資源等等。通過一步步debug以後發現問題出如今

solver.net.set_input_arrays

solver在將數據傳送到網絡低端的時候報錯。那麼接下來找到python目錄下的caffe\python\caffe\_caffe.cpp文件，這個文件是基於boost python的，用來將C++的接口導出，供python調用。進一步咱們找到相關函數

void Net_SetInputArrays(Net<Dtype>* net, bp::object data_obj,
    bp::object labels_obj) {
  // check that this network has an input MemoryDataLayer
  shared_ptr<MemoryDataLayer<Dtype> > md_layer =
    boost::dynamic_pointer_cast<MemoryDataLayer<Dtype> >(net->layers()[0]);
  if (!md_layer) {
    throw std::runtime_error("set_input_arrays may only be called if the"
        " first layer is a MemoryDataLayer");
  }
  // check that we were passed appropriately-sized contiguous memory
  PyArrayObject* data_arr =
      reinterpret_cast<PyArrayObject*>(data_obj.ptr());
  PyArrayObject* labels_arr =
      reinterpret_cast<PyArrayObject*>(labels_obj.ptr());
  CheckContiguousArray(data_arr, "data array", md_layer->channels(),
      md_layer->height(), md_layer->width());
  CheckContiguousArray(labels_arr, "labels array", 1, 1, 1);
  if (PyArray_DIMS(data_arr)[0] != PyArray_DIMS(labels_arr)[0]) {
    throw std::runtime_error("data and labels must have the same first"
        " dimension");
  }
  if (PyArray_DIMS(data_arr)[0] % md_layer->batch_size() != 0) {
    throw std::runtime_error("first dimensions of input arrays must be a"
        " multiple of batch size");
  }
  md_layer->Reset(static_cast<Dtype*>(PyArray_DATA(data_arr)),
      static_cast<Dtype*>(PyArray_DATA(labels_arr)),
      PyArray_DIMS(data_arr)[0]);
}

問題就出在了最後的一個語句

md_layer->Reset(static_cast<Dtype*>(PyArray_DATA(data_arr)),
      static_cast<Dtype*>(PyArray_DATA(labels_arr)),
      PyArray_DIMS(data_arr)[0]);

當執行reset MemoryDataLayer的Reset函數時出錯。於此同時在github上也發現了一樣的問題，https://github.com/BVLC/caffe/issues/2334也是由於Python MemoryDataLayer引起的段錯誤。他說到，在裏面把傳入的data和label作要給深拷貝就能夠解決，估計是運行時數據已經被釋放了，只傳了指針引起了段錯誤。

解決方案：
找到caffe\src\layers\memory_data_layer.cpp打開，給Reset函數添加相應的深拷貝代碼

template <typename Dtype>
void MemoryDataLayer<Dtype>::Reset(Dtype* data, Dtype* labels, int n) {
  CHECK(data);
  CHECK(labels);
  CHECK_EQ(n % batch_size_, 0) << "n must be a multiple of batch size";
  // Warn with transformation parameters since a memory array is meant to
  // be generic and no transformations are done with Reset().
  if (this->layer_param_.has_transform_param()) {
    LOG(WARNING) << this->type() << " does not transform array data on Reset()";
  }
  // data_ = data; 將這裏註釋掉，
  // labels_ = labels;

  //如下部分是進行深拷貝
  if(data_)
      delete []data_;
  if(labels_)
      delete [] labels_;
  data_ = new Dtype[n*size_];
  labels_ = new Dtype[n * num_tasks_];

  memcpy(data_, data, sizeof(Dtype)*n*size_);
  memcpy(labels_, labels, sizeof(Dtype) * n * num_tasks_);

  n_ = n;
  pos_ = 0;
}

Ok進行修改以後，回到Caffe的根目錄，執行make all,make test,`make runtest,make pycaffe。從新編譯完成以後，從新運行就行了，繼續開始訓練。