4dd2f9b9520e9bcd1500624a6869cbc43ae79cbc,train.py,,autoencoder,#,185

Before Change

            threads = tf.train.start_queue_runners(sess=sess, coord=coord)

            // Create the savers.
            restore_saver = build_restore_saver(
                [global_step] if ARGS.checkpoint_path == "" else [],
                scopes_to_remove=ARGS.exclude_scopes)
            train_saver, best_saver = build_train_savers([global_step])

            if ARGS.checkpoint_path != "":
                checkpoint = tf.train.latest_checkpoint(ARGS.checkpoint_path)
                if checkpoint:
                    restore_saver.restore(sess, checkpoint)
                else:
                    print("[E] {} not valid".format(ARGS.checkpoint_path))
                    sys.exit(-1)
            elif not ARGS.restart:  // continue from the saved checkpoint
                // restore previous session if exists
                checkpoint = tf.train.latest_checkpoint(LOG_DIR)
                if checkpoint:
                    restore_saver.restore(sess, checkpoint)
                else:
                    print("[I] Unable to restore from checkpoint")

            train_log, validation_log = build_loggers(sess.graph)

            // Extract previous global step value
            old_gs = sess.run(global_step)

After Change

            threads = tf.train.start_queue_runners(sess=sess, coord=coord)

            // Create the savers.
            restore_or_restart(sess, global_step)
            train_saver, best_saver = build_train_savers([global_step])
            train_log, validation_log = build_loggers(sess.graph)

            // Extract previous global step value

Instances