twodlearn.optim module

class twodlearn.optim.ConstantLr(value)

Bases: object

class twodlearn.optim.DataFeeder(feed_train, feed_valid=None)

Bases: object

feed_train()

feed_valid()

stop()

class twodlearn.optim.EarlyStopping(monitor, var_list, logger_path, session, check_func=None, options=None)

Bases: twodlearn.core.common.TdlObject

add_checkpoint

Decorator used to specify methods that have a set of local variables.

These methods consist of an initialization method (speficied with the function given to the decorator) and an execute method.

check_greather(local)

check_lower

Decorator used to specify methods that have a set of local variables.

These methods consist of an initialization method (speficied with the function given to the decorator) and an execute method.

check_progress(step, monitor)

property checkpoints

reset()

restore

Decorator used to specify methods that have a set of local variables.

These methods consist of an initialization method (speficied with the function given to the decorator) and an execute method.

restore_file()

save

Decorator used to specify methods that have a set of local variables.

These methods consist of an initialization method (speficied with the function given to the decorator) and an execute method.

class twodlearn.optim.EarlyStoppingV2(optimizer, objective, minimize=True)

Bases: twodlearn.core.common.TdlObject

property checkpoints

objective

property optimizer

reset()

restore

Decorator used to specify methods that have a set of local variables.

These methods consist of an initialization method (speficied with the function given to the decorator) and an execute method.

property session

class twodlearn.optim.OptimizationManager(session, optimizer=None, step_op=None, monitor_manager=None, n_logging=100, saver=None, options=None, optimizer_op=None)

Bases: object

Performs a standard mini-batch training with validation evaluation

check_nan(step, xp)

Check if the result from the optimizer includes Nan values. :param step: current step of the optimizer. :type step: int :param xp: list of outputs from the optimizer :type xp: list

Returns

True if variables were reset.

Return type

bool

check_progress(step, xp)

Check if progress was made in the last call to the optimizer :param step: current optimizer step. :type step: int :param xp: list of outputs from the training monitors. :type xp: list

Returns

variables were reset.

Return type

bool

run(n_train_steps, feed_train=None, n_valid_steps=1, valid_eval_freq=1, feed_valid=None, monitor_training=True)

run_step(step, ops, feed_dict)

Run a step of the optimizer. :param step: Description of parameter step. :type step: type :param ops: Description of parameter ops. :type ops: type :param feed_dict: Description of parameter feed_dict. :type feed_dict: type

Returns

Description of returned object.

Return type

type

class twodlearn.optim.Optimizer(loss, var_list, session=None, metrics=None, n_logging=100, log_folder=None, options=None, **kargs)

Bases: twodlearn.core.common.TdlModel

check_nan(step, xp)

Check if the result from the optimizer includes Nan values. :param step: current step of the optimizer. :type step: int :param xp: list of outputs from the optimizer :type xp: list

Returns

True if variables were reset.

Return type

bool

check_progress(step, xp)

Check if progress was made in the last call to the optimizer :param step: current optimizer step. :type step: int :param xp: list of outputs from the training monitors. :type xp: list

Returns

variables were reset.

Return type

bool

feed_train()

learning_rate

log_folder

loss_monitor

monitor_manager

optimizer

reset()

Reset optimizer variables (var_optim)

run(n_train_steps, feed_train=None, n_valid_steps=1, valid_eval_freq=1, feed_valid=None, monitor_training=True)

run_step(step, ops, feed_dict)

Run a step of the optimizer. :param step: Description of parameter step. :type step: type :param ops: Description of parameter ops. :type ops: type :param feed_dict: Description of parameter feed_dict. :type feed_dict: type

Returns

Description of returned object.

Return type

type

saver

session

step_op

property var_optim

Variables created by the optimizer

class twodlearn.optim.SimpleSaver(var_list, logger_path, session)

Bases: twodlearn.core.common.TdlObject

add_checkpoint(step)

property checkpoints

reset()

restore_file()

save

Decorator used to specify methods that have a set of local variables.

These methods consist of an initialization method (speficied with the function given to the decorator) and an execute method.