CODEWITHSUNDEEP
pythontensorflow
bsky
bsky

Reputation: 20222

AttributeError: 'str' object has no attribute 'queue_ref'

I'm trying to read a CSV file using Tensorflow:

import tensorflow as tf

reader = tf.TextLineReader()
key, value = reader.read("../input/training.csv")

However, I get this issue on the last line of the code:

/opt/conda/lib/python3.6/site-packages/tensorflow/python/ops/io_ops.py in read(self, queue, name)
    191       queue_ref = queue
    192     else:
--> 193       queue_ref = queue.queue_ref
    194     if self._reader_ref.dtype == dtypes.resource:
    195       return gen_io_ops._reader_read_v2(self._reader_ref, queue_ref, name=name)

AttributeError: 'str' object has no attribute 'queue_ref'

Any idea what could be the cause of this?

Upvotes: 0

Views: 1166

Answers (1)

Martijn Pieters
Martijn Pieters

Reputation: 1121972

You need to create a queue for your files:

filename_queue = tf.train.string_input_producer(["../input/training.csv"])

reader = tf.TextLineReader()
key, value = reader.read(filename_queue)

From the tf.TextLineReader().read() documentation:

read(
      queue,
      name=None
  )

[...]

  • queue: A Queue or a mutable string Tensor representing a handle to a Queue, with string work items.

and from the Reading from files section of the API guide:

A typical pipeline for reading records from files has the following stages:

  1. The list of filenames
  2. Optional filename shuffling
  3. Optional epoch limit
  4. Filename queue
  5. A Reader for the file format
  6. A decoder for a record read by the reader
  7. Optional preprocessing
  8. Example queue

The tf.train.string_input_producer() call above creates the filename queue from item 4, passing in a simple list of filenames (item 1). The tf.TextLineReader() is item 5 on the above list.

Upvotes: 2

Related Questions