Tensorflow codes exceptionally exit when using thread in a loop -
i describe problem as possible receiving valuable answers. this, posting pseudocode below , pointing out further question:
for _ in range(100): tf.reset_default_graph() read data queue build graph tf.session() sess: coord = tf.train.coordinator() threads = tf.train.start_queue_runners(sess, coord) try: in range(total_steps): if coord.should_stop(): work... except tf.errors.outofrangeerror: print('done') finally: coord.request_stop() coord.join(threads)
in summary, put code in loop, , in each loop. graph reset, , use thread read data.
i have used command "nohup python myscript.py >./log.txt &", while no exceptional or error messages in log.txt.
now, thinking whether threads causing exceptional exit? not sure.
many thanks!
Comments
Post a Comment