TFrecords reader class

tefla.dataset.reader.Reader (dataset, reader_kwargs=None, shuffle=True, num_readers=16, capacity=1, num_epochs=None)

Args

• dataset: an instance of the dataset class
• reader_kwargs: extra arguments to be passed to the TFRecordReader
• shuffle: whether to shuffle the dataset
• num_readers:a int, num of readers to launch
• capacity: a int, capacity of the queue used
• num_epochs: a int, num of epochs for training or validation

Methods

parallel_reader (min_queue_examples=1024)

Primarily used for Training ops

Args

• min_queue_examples: min number of queue examples after dequeue

single_reader (num_epochs=1, shuffle=False, capacity=1)

Data will be read using single TFRecordReader, primarily used for validation

Args

• num_epochs: number of epoch
• shuffle: shuffle the dataset. False for validation
• capacity: queue capacity

Returns

a single item from the tfrecord files

---