uimnet/workers/evaluator.py [122:164]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    loaders_kwargs = dict(
      batch_size=train_cfg.dataset.batch_size,
      shuffle=False,
      pin_memory=True if 'cuda' in eval_cfg.experiment.device else False,
      num_workers=eval_cfg.experiment.num_workers)
    loaders, datanodes = get_loaders_datanodes(partitions, train_cfg,
                                               loaders_kwargs=loaders_kwargs,
                                               seed=eval_cfg.dataset.seed)
    for datanode in datanodes.values():
      datanode.eval()

    num_classes = partitions[('train', 'in')].num_classes


    ###############
    ## Algorithm ##
    ###############
    self.algorithm = Algorithm(num_classes=num_classes,
                               arch=train_cfg.algorithm.arch,
                               device=eval_cfg.experiment.device,
                               use_mixed_precision=train_cfg.algorithm.use_mixed_precision,
                               seed=train_cfg.algorithm.seed,
                               sn=train_cfg.algorithm.sn,
                               sn_coef=train_cfg.algorithm.sn_coef,
                               sn_bn=train_cfg.algorithm.sn_bn
                               )


    utils.message(eval_cfg)
    self.algorithm.initialize()
    self.algorithm.load_state(train_cfg.output_dir,
                              map_location=eval_cfg.experiment.device)

    records = []
    self.algorithm.eval()
    with torch.no_grad():
      for temperature_mode in ['initial', 'learned']:

        self.algorithm.set_temperature(temperature_mode)
        measure = Measure(algorithm=self.algorithm)
        measure.estimate(loaders[('train', 'in')])

        measurements = collections.defaultdict(list)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



uimnet/workers/evaluator2.py [134:176]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    loaders_kwargs = dict(
      batch_size=train_cfg.dataset.batch_size,
      shuffle=False,
      pin_memory=True if 'cuda' in eval_cfg.experiment.device else False,
      num_workers=eval_cfg.experiment.num_workers)
    loaders, datanodes = get_loaders_datanodes(partitions, train_cfg,
                                               loaders_kwargs=loaders_kwargs,
                                               seed=eval_cfg.dataset.seed)
    for datanode in datanodes.values():
      datanode.eval()

    num_classes = partitions[('train', 'in')].num_classes


    ###############
    ## Algorithm ##
    ###############
    self.algorithm = Algorithm(num_classes=num_classes,
                               arch=train_cfg.algorithm.arch,
                               device=eval_cfg.experiment.device,
                               use_mixed_precision=train_cfg.algorithm.use_mixed_precision,
                               seed=train_cfg.algorithm.seed,
                               sn=train_cfg.algorithm.sn,
                               sn_coef=train_cfg.algorithm.sn_coef,
                               sn_bn=train_cfg.algorithm.sn_bn
                               )


    utils.message(eval_cfg)
    self.algorithm.initialize()
    self.algorithm.load_state(train_cfg.output_dir,
                              map_location=eval_cfg.experiment.device)

    records = []
    self.algorithm.eval()
    with torch.no_grad():
      for temperature_mode in ['initial', 'learned']:

        self.algorithm.set_temperature(temperature_mode)
        measure = Measure(algorithm=self.algorithm)
        measure.estimate(loaders[('train', 'in')])

        measurements = collections.defaultdict(list)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -