def __init__(self,
searcher,
dispatcher=None,
callbacks=[],
reward_metric=None,
max_model_size=0,
cache_preprocessed_data=False,
cache_home=None,
**config_kwargs):
self.config_kwargs = config_kwargs
metrics = config_kwargs.get('metrics')
if metrics is None and reward_metric is None:
raise ValueError('Must specify `reward_metric` or `metrics`.')
if reward_metric is None:
reward_metric = metrics[0]
if metrics is None:
metrics = [reward_metric]
config_kwargs['metrics'] = metrics
if reward_metric not in metrics:
metrics.append(reward_metric)
config_kwargs['metrics'] = metrics
self.cache_preprocessed_data = cache_preprocessed_data
self.cache_home = cache_home
HyperModel.__init__(self,
searcher,
dispatcher=dispatcher,
callbacks=callbacks,
reward_metric=reward_metric)
def __init__(self,
searcher,
optimizer,
loss,
metrics,
dispatcher=None,
callbacks=[],
reward_metric=None,
max_model_size=0,
one_shot_mode=False,
one_shot_train_sampler=None,
visualization=False):
self.optimizer = optimizer
self.loss = loss
self.metrics = metrics
self.max_model_size = max_model_size
if reward_metric is None:
reward_metric = metrics[0]
if one_shot_mode:
self.weights_cache = LayerWeightsCache()
else:
self.weights_cache = None
self.one_shot_mode = one_shot_mode
self.one_shot_train_sampler = one_shot_train_sampler if one_shot_train_sampler is not None else searcher
self.visualization = visualization
HyperModel.__init__(self,
searcher,
dispatcher=dispatcher,
callbacks=callbacks,
reward_metric=reward_metric)
def __init__(self,
searcher,
task='classification',
dispatcher=None,
callbacks=None,
reward_metric='accuracy',
data_cleaner_params=None,
cache_dir=None,
clear_cache=True):
if callbacks is None:
callbacks = []
self.task = task
self.data_cleaner_params = data_cleaner_params
self.cache_dir = cache_dir
self.clear_cache = clear_cache
HyperModel.__init__(self,
searcher,
dispatcher=dispatcher,
callbacks=callbacks,
reward_metric=reward_metric)