paddle.fluid.layers.control_flow. array_read ( array, i ) [source]

This OP is used to read data at the specified position from the input array api_fluid_LoDTensorArray . array is the input array and i is the specified read position. This OP is often used together with api_fluid_layers_array_write OP.

Case 1:

System Message: ERROR/3 (/usr/local/lib/python3.8/site-packages/paddle/fluid/layers/ of paddle.fluid.layers.control_flow.array_read, line 8)

Unexpected indentation.

    The shape of first three tensors are [1], and that of the last one is [1,2]:
        array = ([0.6], [0.1], [0.3], [0.4, 0.2])
        i = [3]

    output = [0.4, 0.2]
  • array (LoDTensorArray) – The input LoDTensorArray.

  • i (Variable) – 1-D Tensor, whose shape is [1] and dtype is int64. It represents the specified read position of array.


The LoDTensor or Tensor that is read at the specified position of array.

Return type



# First we're going to create a LoDTensorArray, then we're going to write the Tensor into
# the specified position, and finally we're going to read the Tensor at that position.
import paddle.fluid as fluid
arr = fluid.layers.create_array(dtype='float32')
tmp = fluid.layers.fill_constant(shape=[3, 2], dtype='int64', value=5)
i = fluid.layers.fill_constant(shape=[1], dtype='int64', value=10)
# tmp is the Tensor with shape [3,2], and if we write it into the position with subscript 10
# of the empty-array: arr, then the length of arr becomes 11.
arr = fluid.layers.array_write(tmp, i, array=arr)
# Read the data of the position with subscript 10.
item = fluid.layers.array_read(arr, i)

# You can print out the data via executor.
input = fluid.layers.Print(item, message="The LoDTensor of the i-th position:")
main_program = fluid.default_main_program()
exe = fluid.Executor(fluid.CPUPlace())

# The printed result is:

# 1569588169  The LoDTensor of the i-th position: The place is:CPUPlace
# Tensor[array_read_0.tmp_0]
#    shape: [3,2,]
#    dtype: l
#    data: 5,5,5,5,5,5,

# the output is 2-D Tensor with shape [3,2].
# dtype is the corresponding C++ data type, which may vary in different environments.
# Eg: if the data type of tensor is int64, then the corresponding C++ data type is int64_t,
#       so the dtype value is typeid(int64_t).Name(), which is 'x' on MacOS, 'l' on Linux,
#       and '__int64' on Windows. They both represent 64-bit integer variables.