eval_retrieval_feature_extract.py [92:105]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    )
    model = resnet50(**model_config_dict).cuda()
    model = DistributedDataParallel(model, device_ids=[args.local_rank])
    load_pretrained(args, model)
    model.eval()

    logger.info('model init done')

    all_feat = []
    all_feat_cls = np.zeros([len(data_loader)], dtype=np.int32)

    with torch.no_grad():
        for idx, (data, cls) in enumerate(data_loader):
            logger.info('{}/{}'.format(idx, len(data_loader)))
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



eval_svm_feature_extract.py [86:98]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    )
    model = resnet50(**model_config_dict).cuda()
    model = DistributedDataParallel(model, device_ids=[args.local_rank])
    load_pretrained(args, model)
    model.eval()

    logger.info('model init done')
    # all_feat = np.zeros([len(data_loader), 2048], dtype=np.float32)
    all_feat = []
    all_feat_cls = np.zeros([len(data_loader)], dtype=np.int32)
    with torch.no_grad():
        for idx, (data, cls) in enumerate(data_loader):
            logger.info('{}/{}'.format(idx, len(data_loader)))
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -