Created by: alexqdh

python 2.7 paddle 1.3

每次训练都会重新读取一遍训练，想先把训练数据load到内存，然后每次训练的时候都重新创建一个reader，报了一个错:

EnforceNotMet: LoDTensorBlockingQueueHolder::InitOnce() can only be called once at [/paddle/paddle/fluid/operators/reader/lod_tensor_blocking_queue.h:73]

代码:

    def build_input(self, image_shape, is_train):
        name = 'train_reader' if is_train else 'test_reader'
        py_reader = fluid.layers.py_reader(
            capacity=64,
            shapes=[[-1] + image_shape, [-1, 1]],
            lod_levels=[0, 0],
            dtypes=["float32", "int64"],
            use_double_buffer=True,
            name=name)
        return py_reader