aef60c84d7ad8b5fb852f2dec16b2a8de838a321,engine/training.py,,run,#Any#Any#Any#Any#,24

Before Change

        // tensorboard would like to find the logs from different training runs in separate subdirs
        log_sub_dirs = [name for name in os.listdir(os.path.join(root_dir, "logs"))
                        if os.path.isdir(os.path.join(root_dir, "logs", name))]
        if log_sub_dirs:
            try:
                previous_sub_dir = max([int(name) for name in log_sub_dirs])
            except:
                quit("Any immediate sub directories of the log directory must be integer-numbered")
            if param.starting_iter > 0:
                current_log_sub_dir = str(previous_sub_dir)
            else:
                current_log_sub_dir = str(1 + previous_sub_dir)
        else:
            current_log_sub_dir = "0"
        writer = tf.summary.FileWriter(os.path.join(root_dir, "logs", current_log_sub_dir),
                                       sess.graph)
        try:
            print("Filling the queue (this can take a few minutes)")

After Change

        coord = tf.train.Coordinator()
        // Place logs from each new training run in a new folder for better tensorboard visualization
        log_sub_dirs = [name for name in os.listdir(os.path.join(root_dir, "logs")) if name.isdecimal()]
        if log_sub_dirs and param.starting_iter==0:
            log_sub_dir = str(max([int(name) for name in log_sub_dirs])+1)
        elif log_sub_dirs and param.starting_iter > 0:
            log_sub_dir = str(max([int(name) for name in log_sub_dirs if os.path.isdir(os.path.join(root_dir, "logs", name))]))
        else:
            log_sub_dir = "0"
        writer = tf.summary.FileWriter(os.path.join(root_dir, "logs", log_sub_dir),
                                       sess.graph)
        try:
            print("Filling the queue (this can take a few minutes)")

Instances