def learning_updates(self):
"""
Return updates in the training.
"""
params = self.network.parameters
gradients = T.grad(self.cost, params)
return optimize_updates(params, gradients, self.config)
def optimization_updates(self, params, gradients):
"""
Return updates from optimization.
"""
updates, free_parameters = optimize_updates(params, gradients, self.config)
self.network.free_parameters.extend(free_parameters)
logging.info("Added %d free parameters for optimization" % len(free_parameters))
return updates
def optimization_updates(self, params, gradients):
"""
Return updates from optimization.
"""
updates, free_parameters = optimize_updates(params, gradients, self.config)
self.network.free_parameters.extend(free_parameters)
logging.info("Added %d free parameters for optimization" % len(free_parameters))
return updates
def learning_updates(self):
"""
Return updates in the training.
"""
params = self.network.parameters
gradients = T.grad(self.cost, params)
updates, free_parameters = optimize_updates(params, gradients, self.config)
self.network.free_parameters.extend(free_parameters)
logging.info("Added %d free parameters for optimization" % len(free_parameters))
return updates
def learning_updates(self):
"""
Return updates in the training.
"""
params = self.network.parameters
# Freeze parameters
if self.config.freeze_params:
logging.info("freeze parameters: %s" % ", ".join(map(str, self.config.freeze_params)))
params = [p for p in params if p not in self.config.freeze_params]
gradients = T.grad(self.cost, params)
updates, free_parameters = optimize_updates(params, gradients, self.config)
self.network.free_parameters.extend(free_parameters)
logging.info("Added %d free parameters for optimization" % len(free_parameters))
return updates
def learning_updates(self):
"""
Return updates in the training.
"""
params = self.network.parameters
# Freeze parameters
if self.config.freeze_params:
logging.info("freeze parameters: %s" %
", ".join(map(str, self.config.freeze_params)))
params = [p for p in params if p not in self.config.freeze_params]
gradients = T.grad(self.cost, params)
updates, free_parameters = optimize_updates(params, gradients,
self.config)
self.network.free_parameters.extend(free_parameters)
logging.info("Added %d free parameters for optimization" %
len(free_parameters))
return updates
