"reflection": [0, 0, 0]
}, #Bernoulli p
interp_order=1),
DefaultNormalizer(tags=["bcolzall:3d"])
]
#####################
# training #
#####################
"This is the train dataloader. We will train until this one stops loading data."
"You can set the number of epochs, the datasets and if you want it multiprocessed"
n_epochs = 1000
training_data = BcolzAllDataLoader(sets=TRAINING,
epochs=n_epochs,
preprocessors=preprocessors,
multiprocess=False,
crash_on_exception=True)
"Schedule the reducing of the learning rate. On indexing with the number of epochs, it should return a value for the learning rate."
lr = 0.01
lr_decay = 0.5
learning_rate_schedule = {0.: lr}
for i in range(10):
learning_rate_schedule[float(2**i)] = lr * (lr_decay**(i + 1))
print learning_rate_schedule
"The function to build updates."
build_updates = lasagne.updates.adam
norm_patch_shape=norm_patch_shape,
interp_order=1),
DefaultNormalizer(tags=["bcolzall:3d"])
]
#####################
# training #
#####################
"This is the train dataloader. We will train until this one stops loading data."
"You can set the number of epochs, the datasets and if you want it multiprocessed"
n_epochs = 100
training_data = BcolzAllDataLoader(
sets=TRAINING,
epochs=n_epochs,
preprocessors=preprocessors,
multiprocess=multiprocessing_on,
crash_on_exception=True)
"Schedule the reducing of the learning rate. On indexing with the number of epochs, it should return a value for the learning rate."
lr = 0.00001 * batch_size
lr_min = lr/1000.
lr_decay = 0.9
learning_rate_schedule = {}
for i in range(n_epochs):
lr_ = lr*(lr_decay**i)
if lr_ < lr_min: break
learning_rate_schedule[i] = lr_
print learning_rate_schedule