def dist_init()

in distributed_training/src_dir/dis_util.py [0:0]

• 32 lines of code
• 11 McCabe index (conditional complexity)

def dist_init(fn, args):
    
    if args.seed is not None:
        random.seed(args.seed)
        torch.manual_seed(args.seed)
        np.random.seed(args.seed)
        torch.cuda.manual_seed_all(args.seed)
        cudnn.deterministic = True

        if cudnn.deterministic:
            warnings.warn('You have chosen to seed training. '
                          'This will turn on the CUDNN deterministic setting, '
                          'which can slow down your training considerably! '
                          'You may see unexpected behavior when restarting '
                          'from checkpoints.')

    args.is_distributed = len(args.hosts) > 1 and args.backend is not None
    args.is_multigpus = args.num_gpus > 1
    args.multigpus_distributed = (args.is_distributed or args.is_multigpus)

    logger.debug("multigpus_distributed - {}".format(
        args.multigpus_distributed))
    logger.debug("Number of gpus available - {}".format(args.num_gpus))

    if args.multigpus_distributed and args.exp_cnt == 0:
        if args.apex:
            # Initialize the distributed environment.
            mp.spawn(fn, nprocs=args.num_gpus, args=(args, ))
        else:
            if args.data_parallel:
                sdp, DDP = _sdp_import(args)
                sdp.init_process_group() if not sdp.is_initialized() else None    
            elif args.model_parallel:
                smp.init()
            args.exp_cnt = fn(None, args)
            
            
    else:
        args.exp_cnt = fn(0, args)
    return args