Handle to the underlying TensorFlow reader.
Name of this reader.
Name of this reader.
Creates an op that returns the number of records that this reader has produced.
Creates an op that returns the number of records that this reader has produced.
This is the same as the number of read executions that have succeeded.
Name for the created op.
Created op output, which is an INT64
scalar tensor.
Creates an op that returns the number of work units that this reader has finished processing.
Creates an op that returns the number of work units that this reader has finished processing.
Name for the created op.
Created op output, which is an INT64
scalar tensor.
Creates an op that reads the next record (i.e., key-value pair) produced by this reader.
Creates an op that reads the next record (i.e., key-value pair) produced by this reader.
The op will dequeue from the input queue if necessary (e.g., when the reader needs to start reading from a new file since it has finished with the previous file).
Queue to obtain the work units from.
Name for the created op.
Created op outputs as a key-value pair.
Creates an op that reads up to the next numRecords
records (i.e., key-value pairs) produced by this reader.
Creates an op that reads up to the next numRecords
records (i.e., key-value pairs) produced by this reader.
The op will dequeue from the input queue if necessary (e.g., when the reader needs to start reading from a new file since it has finished with the previous file).
Queue to obtain the work units from.
INT64
scalar tensor specifying how many records to read.
Name for the created op.
Created op outputs as a key-value pair of one-dimensional tensors.
Creates an op that restores this reader to its initial clean state.
Creates an op that restores this reader to its initial clean state.
Name for the created op.
Created op.
Class that supports all TensorFlow reader implementations.
Conceptually, readers convert string "work units" into records (i.e., key-value pairs). Typically the "work units" are filenames and the records are extracted from the contents of those files. We want a single record produced per step, but a work unit can correspond to many records.
Therefore we introduce some decoupling using a Queue. The queue contains the work units and the reader dequeues from the queue when it is asked to produce a record (e.g., via its
read
method) but it has already finished the last work unit.