def __init__(self, prm):
self.log = logger.get_logger('factories')
self.prm = prm
self.dataset_name = self.prm.dataset.DATASET_NAME
self.trainer = self.prm.train.train_control.TRAINER
self.architecture = self.prm.network.ARCHITECTURE
self.learning_rate_setter = self.prm.train.train_control.learning_rate_setter.LEARNING_RATE_SETTER
self.tester = self.prm.test.test_control.TESTER
self.active_selection_criterion = self.prm.train.train_control.ACTIVE_SELECTION_CRITERION
def __init__(self, name, prm, fixed_centers, *args, **kwargs):
super(AllCentersKMeans, self).__init__(*args, **kwargs)
self.name = name
self.prm = prm
self.log = logger.get_logger(name)
self.fixed_centers = fixed_centers
self.n_fixed = fixed_centers.shape[0]
self.init = 'random'
self.n_init = 10
self.verbose = False
self.assert_config()
def __init__(self, name, prm, model, retention):
self.name = name
self.prm = prm
self.model = model
self.retention = retention
self.log = logger.get_logger(name)
self._init_mm = self.prm.network.optimization.DML_MARGIN_MULTIPLIER
self.decay_refractory_steps = self.prm.train.train_control.margin_multiplier_setter.MM_DECAY_REFRACTORY_STEPS
self.global_step_of_last_decay = 0
self._mm = self._init_mm
def __init__(self, name, prm, model, steps_to_save, checkpoint_dir, saver, checkpoint_basename='model_schedule.ckpt'):
self.name = name
self.prm = prm
self.log = logger.get_logger(name)
self.model = model # model might change between runs, cannot use global train step. Must use model step.
self._saver = saver
self._checkpoint_dir = checkpoint_dir
self._save_path = os.path.join(checkpoint_dir, checkpoint_basename)
if steps_to_save is None:
steps_to_save = []
self._steps_to_save = steps_to_save
def __init__(self, name, prm, fixed_centers, random_state, *args,
**kwargs):
super(KMeansWrapper, self).__init__(*args, **kwargs)
self.name = name
self.prm = prm
self.log = logger.get_logger(name)
self.fixed_centers = fixed_centers
self.n_fixed = fixed_centers.shape[0]
self.random_state = random_state
self.init = 'random'
self.n_init = 1
self.verbose = True
self.assert_config()
def __init__(self, name, prm, model, retention):
self.name = name
self.prm = prm
self.model = model
self.retention = retention # used in children
self.log = logger.get_logger(name)
self.learning_rate_setter = self.prm.train.train_control.learning_rate_setter.LEARNING_RATE_SETTER
self._init_lrn_rate = self.prm.network.optimization.LEARNING_RATE
self._reset_lrn_rate = self.prm.train.train_control.learning_rate_setter.LEARNING_RATE_RESET
if self._reset_lrn_rate is None:
self.log.warning(
'LEARNING_RATE_RESET is None. Setting LEARNING_RATE_RESET=LEARNING_RATE'
)
self._reset_lrn_rate = self.prm.network.optimization.LEARNING_RATE
self._lrn_rate = self._init_lrn_rate
def __init__(self, name, prm):
super(DatasetWrapper, self).__init__(name)
self.prm = prm
self.log = logger.get_logger(name)
self.dataset_name = self.prm.dataset.DATASET_NAME
self.train_set_size = self.prm.dataset.TRAIN_SET_SIZE
self.validation_set_size = self.prm.dataset.VALIDATION_SET_SIZE
self.test_set_size = self.prm.dataset.TEST_SET_SIZE
self.train_validation_map_ref = self.prm.dataset.TRAIN_VALIDATION_MAP_REF
self.H = self.prm.network.IMAGE_HEIGHT
self.W = self.prm.network.IMAGE_WIDTH
self.train_batch_size = self.prm.train.train_control.TRAIN_BATCH_SIZE
self.eval_batch_size = self.prm.train.train_control.EVAL_BATCH_SIZE
self.rand_gen = np.random.RandomState(prm.SUPERSEED)
self.train_validation_info = []
self.train_dataset = None
self.train_eval_dataset = None
self.validation_dataset = None
self.test_dataset = None
self.iterator = None
self.train_iterator = None # static iterator for train only
self.train_eval_iterator = None # dynamic iterator for train evaluation. need to reinitialize
self.validation_iterator = None # dynamic iterator for validation. need to reinitialize
self.test_iterator = None # dynamic iterator for test. need to reinitialize
self.handle = None
self.train_handle = None
self.train_eval_handle = None
self.validation_handle = None
self.test_handle = None
self.next_minibatch = None # this is the output of iterator.get_next()
if self.validation_set_size is None:
self.log.warning('Validation set size is None. Setting its size to 0')
self.validation_set_size = 0
self.train_validation_size = self.train_set_size + self.validation_set_size
import lib.logger.logger as logger
from lib.logger.logging_config import logging_config
from utils.parameters import Parameters
from utils.factories import Factories
import numpy as np
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
from lib.active_kmean import KMeansWrapper
from sklearn.datasets import make_blobs
logging = logging_config()
logging.disable(logging.DEBUG)
log = logger.get_logger('main')
prm_file = '/data/gilad/logs/log_2210_220817_wrn-fc2_kmeans_SGD_init_200_clusters_4_cap_204/parameters.ini'
prm = Parameters()
prm.override(prm_file)
dev = prm.network.DEVICE
factories = Factories(prm)
model = factories.get_model()
model.print_stats() # debug
preprocessor = factories.get_preprocessor()
preprocessor.print_stats() # debug
def __init__(self, name):
self.name = name
self.log = logger.get_logger(name)
def __init__(self):
self.log = logger.get_logger('FrozenIniParser')
super(FrozenClass, self).__init__()
def __init__(self):
self.log = logger.get_logger('IniParser')
def __init__(self, name, prm, model, *args, **kwargs):
super(TrainSummarySaverHook, self).__init__(*args, **kwargs)
self.name = name
self.prm = prm
self.log = logger.get_logger(name)
self.model = model