tensorflow - SyncReplicasOptimizer Not Looping Over All Epochs -

after running following code, global_step variable not equal num_batches * epochs expected. instead, stops @ end of 1 epoch. asynchronous code without syncreplicasoptimiser works expected though. there i'm doing wrong? here's code:

(args.steps more num_batches * epochs)

        sess = sv.prepare_or_wait_for_session(server.target)         queue_runners = tf.get_collection(tf.graphkeys.queue_runners)         sv.start_queue_runners(sess, queue_runners)          tf.logging.info('started %d queues processing input data.',                         len(queue_runners))          if is_chief:                 sv.start_queue_runners(sess, chief_queue_runners)                 sess.run(init_tokens_op)          print("{0} session ready".format(datetime.now().isoformat()))         #####################################################################          ########################### training loop ###########################         _current_state = np.zeros((batch_size, state_size))         batch_idx in range(args.steps):             if sv.should_stop() or tf_feed.should_stop():                 break              batchx, batchy = feed_dict(tf_feed.next_batch(batch_size))              print('==========================================================')             print(_current_state)              if args.mode == "train":                 _total_loss, _train_step, _current_state, _predictions_series, _global_step = sess.run(                 [total_loss, train_step, current_state, predictions_series, global_step],                 feed_dict={                     batchx_placeholder:batchx,                     batchy_placeholder:batchy,                     init_state:_current_state                 })                  print(_global_step, batch_idx)                 print(_current_state)                 print('==========================================================')                  if _global_step % 5 == 0:                     print("step", _global_step, "loss", _total_loss)                           # else: # todo code checking             #################################################################                  if sv.should_stop() or batch_idx > args.steps:                     tf_feed.terminate()          print("{0} stopping supervisor".format(datetime.now().isoformat()))         sv.stop()