Function main

main_simmim_pt.py:70–117

(config)

def main(config):
    data_loader_train = build_loader(config, simmim=True, is_pretrain=True)

    logger.info(f"Creating model:{config.MODEL.TYPE}/{config.MODEL.NAME}")
    model = build_model(config, is_pretrain=True)
    model.cuda()
    logger.info(str(model))

    optimizer = build_optimizer(config, model, simmim=True, is_pretrain=True)
    model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[config.LOCAL_RANK], broadcast_buffers=False)
    model_without_ddp = model.module

    n_parameters = sum(p.numel() for p in model.parameters() if p.requires_grad)
    logger.info(f"number of params: {n_parameters}")
    if hasattr(model_without_ddp, 'flops'):
        flops = model_without_ddp.flops()
        logger.info(f"number of GFLOPs: {flops / 1e9}")

    lr_scheduler = build_scheduler(config, optimizer, len(data_loader_train))
    scaler = amp.GradScaler()

    if config.TRAIN.AUTO_RESUME:
        resume_file = auto_resume_helper(config.OUTPUT, logger)
        if resume_file:
            if config.MODEL.RESUME:
                logger.warning(f"auto-resume changing resume file from {config.MODEL.RESUME} to {resume_file}")
            config.defrost()
            config.MODEL.RESUME = resume_file
            config.freeze()
            logger.info(f'auto resuming from {resume_file}')
        else:
            logger.info(f'no checkpoint found in {config.OUTPUT}, ignoring auto resume')

    if config.MODEL.RESUME:
        load_checkpoint(config, model_without_ddp, optimizer, lr_scheduler, scaler, logger)

    logger.info("Start training")
    start_time = time.time()
    for epoch in range(config.TRAIN.START_EPOCH, config.TRAIN.EPOCHS):
        data_loader_train.sampler.set_epoch(epoch)

        train_one_epoch(config, model, data_loader_train, optimizer, epoch, lr_scheduler, scaler)
        if dist.get_rank() == 0 and (epoch % config.SAVE_FREQ == 0 or epoch == (config.TRAIN.EPOCHS - 1)):
            save_checkpoint(config, epoch, model_without_ddp, 0., optimizer, lr_scheduler, scaler, logger)

    total_time = time.time() - start_time
    total_time_str = str(datetime.timedelta(seconds=int(total_time)))
    logger.info('Training time {}'.format(total_time_str))


def train_one_epoch(config, model, data_loader, optimizer, epoch, lr_scheduler, scaler):