INSTRUCTION
stringlengths
1
8.43k
RESPONSE
stringlengths
75
104k
General optimization function for Theano. Parameters: params - parameters gradients - gradients config - training config Returns: Theano updates: type config: deepy. TrainerConfig or dict
def optimize_updates(params, gradients, config=None, shapes=None): """ General optimization function for Theano. Parameters: params - parameters gradients - gradients config - training config Returns: Theano updates :type config: deepy.TrainerConfig or dict """ ...
Create a optimizing function receives gradients. Parameters: params - parameters config - training configuration Returns: updating function receives gradients
def optimize_function(params, config=None): """ Create a optimizing function receives gradients. Parameters: params - parameters config - training configuration Returns: updating function receives gradients """ gs = [dim_to_var(p.ndim) for p in params] updates, _ = op...
Return updates in the training.
def _learning_updates(self): """ Return updates in the training. """ params = self.training_params() gradients = self.get_gradients(params) return self.optimization_updates(params, gradients)
Get parameters to be optimized.
def training_params(self): """ Get parameters to be optimized. """ params = self.network.parameters # Freeze parameters if self.config.fixed_parameters: logging.info("fixed parameters: %s" % ", ".join(map(str, self.config.fixed_parameters))) params...
Return updates from optimization.
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 optimiz...
Get the learning function.: param func:: return:
def learning_function(self): """ Get the learning function. :param func: :return: """ network_updates = list(self.network.updates) + list(self.network.training_updates) learning_updates = list(self._learning_updates()) update_list = network_updates + learn...
Parameters: x_t - 28x28 image l_p - 2x1 focus vector Returns: 4x12 matrix
def _glimpse_sensor(self, x_t, l_p): """ Parameters: x_t - 28x28 image l_p - 2x1 focus vector Returns: 4x12 matrix """ # Turn l_p to the left-top point of rectangle l_p = l_p * 14 + 14 - 2 l_p = T.cast(T.round(l_p), "int32") ...
Parameters: x_t - 28x28 image l_p - 2x1 focus vector Returns: 7 * 14 matrix
def _refined_glimpse_sensor(self, x_t, l_p): """ Parameters: x_t - 28x28 image l_p - 2x1 focus vector Returns: 7*14 matrix """ # Turn l_p to the left-top point of rectangle l_p = l_p * 14 + 14 - 4 l_p = T.cast(T.round(l_p), "int...
Parameters: x_t - 28x28 image l_p - 2x1 focus vector Returns: 4x12 matrix
def _glimpse_network(self, x_t, l_p): """ Parameters: x_t - 28x28 image l_p - 2x1 focus vector Returns: 4x12 matrix """ sensor_output = self._refined_glimpse_sensor(x_t, l_p) sensor_output = T.flatten(sensor_output) h_g = self._...
Parameters: h_t - 256x1 vector Returns: 10x1 vector
def _action_network(self, h_t): """ Parameters: h_t - 256x1 vector Returns: 10x1 vector """ z = self._relu(T.dot(h_t, self.W_a) + self.B_a) return self._softmax(z)
Get baseline model. Parameters: model - model path Returns: network
def get_network(model=None, std=0.005, disable_reinforce=False, random_glimpse=False): """ Get baseline model. Parameters: model - model path Returns: network """ network = NeuralClassifier(input_dim=28 * 28) network.stack_layer(FirstGlimpseLayer(std=std, disable_reinforce=di...
Compute first glimpse position using down - sampled image.
def _first_glimpse_sensor(self, x_t): """ Compute first glimpse position using down-sampled image. """ downsampled_img = theano.tensor.signal.downsample.max_pool_2d(x_t, (4,4)) downsampled_img = downsampled_img.flatten() first_l = T.dot(downsampled_img, self.W_f) ...
Parameters: x_t - 28x28 image l_p - 2x1 focus vector h_p - 256x1 vector Returns: h_t 256x1 vector
def _core_network(self, l_p, h_p, x_t): """ Parameters: x_t - 28x28 image l_p - 2x1 focus vector h_p - 256x1 vector Returns: h_t, 256x1 vector """ g_t = self._glimpse_network(x_t, l_p) h_t = self._tanh(T.dot(g_t, self.W_h_g)...
All codes that create parameters should be put into setup function.
def prepare(self): """ All codes that create parameters should be put into 'setup' function. """ self.output_dim = 10 self.encoder = Chain(self.input_dim).stack(Dense(self.internal_layer_size, 'tanh')) self.decoder = Chain(self.internal_layer_size).stack(Dense(self.input_...
Build the computation graph here.
def compute_tensor(self, x): """ Build the computation graph here. """ internal_variable = self.encoder.compute_tensor(x) decoding_output = self.decoder.compute_tensor(internal_variable) classification_output = self.classifier.compute_tensor(internal_variable) ...
Process all data with given function. The scheme of function should be x y - > x y.
def map(self, func): """ Process all data with given function. The scheme of function should be x,y -> x,y. """ if self._train_set: self._train_set = map(func, self._train_set) if self._valid_set: self._valid_set = map(func, self._valid_set) ...
Make targets be one - hot vectors.
def vectorize_target(self, size): """ Make targets be one-hot vectors. """ if self._train_set: self._train_set = self._vectorize_set(self._train_set, size) if self._valid_set: self._valid_set = self._vectorize_set(self._valid_set, size) if self._te...
Print dataset statistics.
def report(self): """ Print dataset statistics. """ logging.info("%s train=%d valid=%d test=%d" % (self.__class__.__name__, len(list(self._train_set)) if self._train_set else 0, ...
We train over mini - batches and evaluate periodically.
def train(self, train_set, valid_set=None, test_set=None, train_size=None): '''We train over mini-batches and evaluate periodically.''' iteration = 0 while True: if not iteration % self.config.test_frequency and test_set: try: self.test(iteration, ...
Sample outputs from LM.
def sample(self, input, steps): """ Sample outputs from LM. """ inputs = [[onehot(self.input_dim, x) for x in input]] for _ in range(steps): target = self.compute(inputs)[0,-1].argmax() input.append(target) inputs[0].append(onehot(self.input_di...
: param x: ( batch time vec )
def compute_tensor(self, x): """ :param x: (batch, time, vec) """ # Target class class_matrix = self.target_tensor // self.output_size class_vector = class_matrix.reshape((-1,)) # Target index target_matrix = self.target_tensor % self.output_size t...
Compute the alignment weights based on the previous state.
def compute_alignments(self, prev_state, precomputed_values, mask=None): """ Compute the alignment weights based on the previous state. """ WaSp = T.dot(prev_state, self.Wa) UaH = precomputed_values # For test time the UaH will be (time, output_dim) if UaH.ndim =...
Compute the context vector with soft attention.
def compute_context_vector(self, prev_state, inputs, precomputed_values=None, mask=None): """ Compute the context vector with soft attention. """ precomputed_values = precomputed_values if precomputed_values else self.precompute(inputs) align_weights = self.compute_alignments(pre...
Train the model in multi - GPU environment.
def train(self, train_set, valid_set=None, test_set=None, train_size=None): """ Train the model in multi-GPU environment. """ from platoon.channel import Worker from platoon.param_sync import EASGD, ASGD server_port = self._port param_map = self.create_param_map()...
A utility function of concatenate.
def concatenate(vars, axis=-1): """ A utility function of concatenate. """ from deepy.core.neural_var import NeuralVariable if isinstance(vars[0], NeuralVariable): concat_var = Concatenate(axis=axis).compute(*vars) if axis == -1 or axis == vars[0].tensor.ndim - 1: concat_...
Wrap a Theano tensor into the variable for defining neural network.: param last_dim: last dimension of tensor 0 indicates that the last dimension is flexible: rtype: deepy. core. neural_var. NeuralVariable
def var(tensor_type, last_dim=0, test_shape=None): """ Wrap a Theano tensor into the variable for defining neural network. :param last_dim: last dimension of tensor, 0 indicates that the last dimension is flexible :rtype: deepy.core.neural_var.NeuralVariable """ # Create tensor from deepy.co...
Pad sequences to given length in the left or right side.
def _pad(self, side, length): """ Pad sequences to given length in the left or right side. """ if self._train_set: self._train_set = pad_dataset(self._train_set, side, length) if self._valid_set: self._valid_set = pad_dataset(self._valid_set, side, length)...
RMSPROP optimization core.
def rmsprop_core(params, gradients, momentum=0.9, learning_rate=0.01): """ RMSPROP optimization core. """ for param, grad in zip(params, gradients): rms_ = theano.shared(np.zeros_like(param.get_value()), name=param.name + '_rms') rms = momentum * rms_ + (1 - momentum) * grad * gr...
Pad data set to specified length. Parameters: length - max length a just to the max length in the batch if length is - 1
def pad_dataset(subset, side="right", length=-1): """ Pad data set to specified length. Parameters: length - max length, a just to the max length in the batch if length is -1 """ assert length == -1 or length > 0 if type(subset[0][0][0]) in [float, int, np.int64, np.int32, np.float32]: ...
Prepare for one epoch. Returns: bool: False if to stop the training.
def prepare_epoch(self): """ Prepare for one epoch. Returns: bool: False if to stop the training. """ self.epoch += 1 if self.epoch >= self.epoch_start_halving and ((self.epoch - self.epoch_start_halving) % self._halving_freq == 0): self._lr *= 0.5...
Handles a control_request received from a worker. Returns: string or dict: response
def handle_control(self, req, worker_id, req_info): """ Handles a control_request received from a worker. Returns: string or dict: response 'stop' - the worker should quit 'wait' - wait for 1 second 'eval' - evaluate on valid and test set to start...
Report elapsed time.
def report(self): """ Report elapsed time. """ if not self.end_time: self.end() print ("Time: {} mins".format((self.end_time - self.start_time )/ 60))
Compare to previous records and return whether the given cost is a new best.: return: True if the given cost is a new best
def compare(self, cost_map): """ Compare to previous records and return whether the given cost is a new best. :return: True if the given cost is a new best """ cri_val = cost_map[self._criteria] if self._best_criteria is None: self._best_criteria = cri_val ...
Run the model with validation data and return costs.
def run(self, data_x): """ Run the model with validation data and return costs. """ output_vars = self.compute(*data_x) return self._extract_costs(output_vars)
This function will be called after each iteration.
def invoke(self): """ This function will be called after each iteration. """ self._counter += 1 if self._counter % self._freq == 0: cnt = 0. sum_map = defaultdict(float) for x in self._trainer.get_data(self._data_split): val_map...
Create inner loop variables.
def _build_loop_vars(self): """ Create inner loop variables. """ from theano.tensor.var import TensorVariable from deepy.core.neural_var import NeuralVariable if not self._loop_vars: self._ordered_out_keys = self._outputs.keys() seq_keys = self._se...
Internal scan with dummy input variables.
def _scan_step(self, vars): """ Internal scan with dummy input variables. """ from neural_var import NeuralVariable if not self._loop_vars: raise Exception("The loop is not initialized. To initialize the loop, use `with loop as vars`") replace_map = {} ...
Get the outputs of the loop. Return specific variables by passing the keys to the arguments.: rtype: MapDict
def get_outputs(self, *args): """ Get the outputs of the loop. Return specific variables by passing the keys to the arguments. :rtype: MapDict """ if args: output_vars = map(self._scan_outputs.get, args) if len(output_vars) == 1: re...
Momentum SGD optimization core.
def momentum_core(params, gradients, momentum=0.9, learning_rate=0.01): """ Momentum SGD optimization core. """ free_parameters = [] updates = [] for param, grad in zip(params, gradients): delta = learning_rate * grad velocity = theano.shared(np.zeros_like(param.get_value...
Execute then_branch when training.
def iftrain(self, then_branch, else_branch): """ Execute `then_branch` when training. """ return ifelse(self._training_flag, then_branch, else_branch, name="iftrain")
Switch training mode.: param flag: switch on training mode when flag is True.
def switch_training(self, flag): """ Switch training mode. :param flag: switch on training mode when flag is True. """ if self._is_training == flag: return self._is_training = flag if flag: self._training_flag.set_value(1) else: sel...
Nesterov s Accelerated Gradient ( NAG ). See http:// www. cs. toronto. edu/ ~fritz/ absps/ momentum. pdf. Still unfinished
def nag_core(params, J, momentum=0.9, learning_rate=0.01): """ Nesterov's Accelerated Gradient (NAG). See http://www.cs.toronto.edu/~fritz/absps/momentum.pdf . Still unfinished """ # TODO: this requires some refractorings. for param in params: step = theano.shared(np.zeros_like(param...
Skip N batches in the training.
def skip(self, n_batches, n_epochs=0): """ Skip N batches in the training. """ logging.info("skip %d epochs and %d batches" % (n_epochs, n_batches)) self._skip_batches = n_batches self._skip_epochs = n_epochs
Load parameters for the training. This method can load free parameters and resume the training progress.
def load_params(self, path, exclude_free_params=False): """ Load parameters for the training. This method can load free parameters and resume the training progress. """ self.network.load_params(path, exclude_free_params=exclude_free_params) self.best_params = self.copy_pa...
Add iteration callbacks function ( receives an argument of the trainer ).: param controllers: can be a TrainingController or a function.: type funcs: list of TrainingContoller
def add_iter_controllers(self, *controllers): """ Add iteration callbacks function (receives an argument of the trainer). :param controllers: can be a `TrainingController` or a function. :type funcs: list of TrainingContoller """ for controller in controllers: ...
Add epoch callbacks function.: param controllers: can be a TrainingController or a function.
def add_epoch_controllers(self, *controllers): """ Add epoch callbacks function. :param controllers: can be a `TrainingController` or a function. """ for controller in controllers: if isinstance(controller, TrainingController): controller.bind(self) ...
Train the model and return costs.
def train(self, train_set, valid_set=None, test_set=None, train_size=None): """ Train the model and return costs. """ self._epoch = 0 while True: if self._skip_epochs > 0: logging.info("skipping one epoch ...") self._skip_epochs -= 1 ...
Run one training iteration.
def _run_train(self, epoch, train_set, train_size=None): """ Run one training iteration. """ self.network.train_logger.record_epoch(epoch + 1) costs = self.train_step(train_set, train_size) if not epoch % self.config.monitor_frequency: self.report(dict(costs),...
Run one valid iteration return true if to continue training.
def _run_valid(self, epoch, valid_set, dry_run=False, save_path=None): """ Run one valid iteration, return true if to continue training. """ costs = self.valid_step(valid_set) # this is the same as: (J_i - J_f) / J_i > min improvement _, J = costs[0] new_best = Fa...
Report the scores and record them in the log.
def report(self, score_map, type="valid", epoch=-1, new_best=False): """ Report the scores and record them in the log. """ type_str = type if len(type_str) < 5: type_str += " " * (5 - len(type_str)) info = " ".join("%s=%.2f" % el for el in score_map.items()) ...
Get specified split of data.
def get_data(self, data_split="train"): """ Get specified split of data. """ if data_split == 'train': return self._current_train_set elif data_split == 'valid': return self._current_valid_set elif data_split == 'test': return self._cur...
Run until the end.: param epoch_controllers: deprecated
def run(self, train_set, valid_set=None, test_set=None, train_size=None, epoch_controllers=None): """ Run until the end. :param epoch_controllers: deprecated """ epoch_controllers = epoch_controllers if epoch_controllers else [] epoch_controllers += self._epoch_controller...
: type bunch_stack: list of list of int
def _cut_to_pieces(self, bunch_stack): """ :type bunch_stack: list of list of int """ stack_len = len(bunch_stack[0]) for i in xrange(0, stack_len, self.fragment_length): yield np.array(map(lambda stack: stack[i: i + self.fragment_length], bunch_stack))
: type bunch_stack: list of list
def _pad_zeros(self, bunch_stack): """ :type bunch_stack: list of list """ min_len = min(map(len, bunch_stack)) for i in range(len(bunch_stack)): bunch_stack[i] = bunch_stack[i][:min_len]
Apply a function to tensors.
def apply(self, func, dim=None): """ Apply a function to tensors. """ output_dim = dim if dim else self.output_dim return NeuralVariable(func(self.tensor), output_dim)
Rprop optimizer. See http:// sci2s. ugr. es/ keel/ pdf/ algorithm/ articulo/ 2003 - Neuro - Igel - IRprop +. pdf.
def rprop_core(params, gradients, rprop_increase=1.01, rprop_decrease=0.99, rprop_min_step=0, rprop_max_step=100, learning_rate=0.01): """ Rprop optimizer. See http://sci2s.ugr.es/keel/pdf/algorithm/articulo/2003-Neuro-Igel-IRprop+.pdf. """ for param, grad in zip(params, gradients): ...
Report usage of training parameters.
def report(self): """ Report usage of training parameters. """ if self.logger: self.logger.info("accessed parameters:") for key in self.used_parameters: self.logger.info(" - %s %s" % (key, "(undefined)" if key in self.undefined_parameters else ""))
Create a parameters block.: param layers: register some layers in the block: param name: specify the name of this block
def new_block(self, *layers, **kwargs): """ Create a parameters block. :param layers: register some layers in the block :param name: specify the name of this block """ from deepy.layers.block import Block block = Block(*layers, **kwargs) return block
An alias of deepy. tensor. var.
def var(self, tensor_type, last_dim=0, test_shape=None): """ An alias of deepy.tensor.var. """ from deepy.tensor import var return var(tensor_type, last_dim=last_dim, test_shape=test_shape)
Create vars given a dataset and set test values. Useful when dataset is already defined.
def create_vars_from_data(self, dataset, split="train"): """ Create vars given a dataset and set test values. Useful when dataset is already defined. """ from deepy.core.neural_var import NeuralVariable vars = [] if split == "valid": data_split = datas...
A loop function the usage is identical with the theano one.: type block: deepy. layers. Block
def scan(self, func, sequences=None, outputs=None, non_sequences=None, block=None, **kwargs): """ A loop function, the usage is identical with the theano one. :type block: deepy.layers.Block """ results, updates = Scanner(func, sequences, outputs, non_sequences, neural_computatio...
Start a loop. Usage: with deepy. graph. loop ( sequences = { x: x } outputs = { o: None } ) as vars: vars. o = vars. x + 1 loop_outputs = deepy. graph. loop_outputs () result = loop_outputs. o
def loop(self, sequences=None, outputs=None, non_sequences=None, block=None, **kwargs): """ Start a loop. Usage: ``` with deepy.graph.loop(sequences={"x": x}, outputs={"o": None}) as vars: vars.o = vars.x + 1 loop_outputs = deepy.graph.loop_outputs() r...
Get a trainer to optimize given model.: rtype: deepy. trainers. GeneralNeuralTrainer
def get_trainer(self, model, method='sgd', config=None, annealer=None, validator=None): """ Get a trainer to optimize given model. :rtype: deepy.trainers.GeneralNeuralTrainer """ from deepy.trainers import GeneralNeuralTrainer return GeneralNeuralTrainer(model, method=me...
Create a shared theano scalar value.
def shared(self, value, name=None): """ Create a shared theano scalar value. """ if type(value) == int: final_value = np.array(value, dtype="int32") elif type(value) == float: final_value = np.array(value, dtype=env.FLOATX) else: final_...
Load parameters from file to fill all blocks sequentially.: type blocks: list of deepy. layers. Block
def fill_parameters(self, path, blocks, exclude_free_params=False, check_parameters=False): """ Load parameters from file to fill all blocks sequentially. :type blocks: list of deepy.layers.Block """ if not os.path.exists(path): raise Exception("model {} does not exis...
Return size of training data. ( optional ): rtype: number
def train_size(self): """ Return size of training data. (optional) :rtype: number """ train_set = self.train_set() if isinstance(train_set, collections.Iterable): return len(list(train_set)) else: return None
Run it return whether to end training.
def invoke(self): """ Run it, return whether to end training. """ self._iter += 1 if self._iter - max(self._trainer.best_iter, self._annealed_iter) >= self._patience: if self._annealed_times >= self._anneal_times: logging.info("ending") ...
Run it return whether to end training.
def invoke(self): """ Run it, return whether to end training. """ self._iter += 1 logging.info("{} epochs left to run".format(self._patience - self._iter)) if self._iter >= self._patience: self._trainer.exit()
Perform reparameterization trick for latent variables.: param layer_size: the size of latent variable
def stack_reparameterization_layer(self, layer_size): """ Perform reparameterization trick for latent variables. :param layer_size: the size of latent variable """ self.rep_layer = ReparameterizationLayer(layer_size, sample=self.sample) self.stack_encoders(self.rep_layer)
Stack encoding layers this must be done before stacking decoding layers.
def stack_encoders(self, *layers): """ Stack encoding layers, this must be done before stacking decoding layers. """ self.stack(*layers) self.encoding_layes.extend(layers)
Stack decoding layers.
def stack_decoders(self, *layers): """ Stack decoding layers. """ self.stack(*layers) self.decoding_layers.extend(layers)
Encode given input.
def encode(self, x): """ Encode given input. """ if not self.encoding_network: self.encoding_network = NeuralNetwork(self.input_dim, self.input_tensor) self.encoding_network.input_variables = self.input_variables for layer in self.encoding_layes: ...
Decode given representation.
def decode(self, x): """ Decode given representation. """ if not self.rep_dim: raise Exception("rep_dim must be set to decode.") if not self.decoding_network: self.decoding_network = NeuralNetwork(self.rep_dim) for layer in self.decoding_layers...
This function creates a 2d gaussian kernel with the standard deviation denoted by sigma
def create_2d_gaussian(dim, sigma): """ This function creates a 2d gaussian kernel with the standard deviation denoted by sigma :param dim: integer denoting a side (1-d) of gaussian kernel :param sigma: floating point indicating the standard deviation :returns: a numpy 2d array """ # ...
This method performs elastic transformations on an image by convolving with a gaussian kernel.: param image: a numpy nd array: kernel_dim: dimension ( 1 - D ) of the gaussian kernel: param sigma: standard deviation of the kernel: param alpha: a multiplicative factor for image after convolution: param negated: a flag in...
def elastic_distortion(image, kernel_dim=21, sigma=6, alpha=30, negated=True): """ This method performs elastic transformations on an image by convolving with a gaussian kernel. :param image: a numpy nd array :kernel_dim: dimension(1-D) of the gaussian kernel :param sigma: standard deviation of ...
Stack a neural layer.: type layer: NeuralLayer: param no_setup: whether the layer is already initialized
def stack_layer(self, layer, no_setup=False): """ Stack a neural layer. :type layer: NeuralLayer :param no_setup: whether the layer is already initialized """ if layer.name: layer.name += "%d" % (len(self.layers) + 1) if not self.layers: la...
Register the layer so that it s param will be trained. But the output of the layer will not be stacked.
def register_layer(self, layer): """ Register the layer so that it's param will be trained. But the output of the layer will not be stacked. """ if type(layer) == Block: layer.fix() self.parameter_count += layer.parameter_count self.parameters.extend(l...
Monitoring the outputs of each layer. Useful for troubleshooting convergence problems.
def monitor_layer_outputs(self): """ Monitoring the outputs of each layer. Useful for troubleshooting convergence problems. """ for layer, hidden in zip(self.layers, self._hidden_outputs): self.training_monitors.append(('mean(%s)' % (layer.name), abs(hidden).mean()))
Return all parameters.
def all_parameters(self): """ Return all parameters. """ params = [] params.extend(self.parameters) params.extend(self.free_parameters) return params
Set up variables.
def setup_variables(self): """ Set up variables. """ if self.input_tensor: if type(self.input_tensor) == int: x = dim_to_var(self.input_tensor, name="x") else: x = self.input_tensor else: x = T.matrix('x') ...
Return network output.
def compute(self, *x): """ Return network output. """ self._compile() outs = self._compute(*x) if self._output_keys: return MapDict(dict(zip(self._output_keys, outs))) else: return outs
Save parameters to file.
def save_params(self, path, new_thread=False): """ Save parameters to file. """ save_logger.info(path) param_variables = self.all_parameters params = [p.get_value().copy() for p in param_variables] if new_thread: thread = Thread(target=save_network_par...
Load parameters from file.
def load_params(self, path, exclude_free_params=False): """ Load parameters from file. """ if not os.path.exists(path): return; logging.info("loading parameters from %s" % path) # Decide which parameters to load if exclude_free_params: params_to_load =...
Print network statistics.
def report(self): """ Print network statistics. """ logging.info("network inputs: %s", " ".join(map(str, self.input_variables))) logging.info("network targets: %s", " ".join(map(str, self.target_variables))) logging.info("network parameters: %s", " ".join(map(str, self.al...
Initialize the layer.: param no_prepare: avoid calling preparation function
def init(self, input_dim=0, input_dims=None, no_prepare=False): """ Initialize the layer. :param no_prepare: avoid calling preparation function """ if self.initialized: return # configure input dimensions if input_dims: self.input_dims = in...
Compute based on NeuralVariable.: type inputs: list of NeuralVariable: return: NeuralVariable
def compute(self, *inputs, **kwargs): """ Compute based on NeuralVariable. :type inputs: list of NeuralVariable :return: NeuralVariable """ from deepy.core.neural_var import NeuralVariable from deepy.core.graph import graph if type(inputs[0]) != NeuralVar...
Let the given block or network manage the parameters of this layer.: param block: Block or NeuralNetwork: return: NeuralLayer
def belongs_to(self, block): """ Let the given block or network manage the parameters of this layer. :param block: Block or NeuralNetwork :return: NeuralLayer """ if self._linked_block: raise SystemError("The layer {} has already blonged to {}".format(self.nam...
Register parameters.
def register_parameters(self, *parameters): """ Register parameters. """ for param in parameters: self.parameter_count += np.prod(param.get_value().shape) self.parameters.extend(parameters)
Register updates that will be executed in each iteration.
def register_updates(self, *updates): """ Register updates that will be executed in each iteration. """ for key, node in updates: if key not in self._registered_updates: self.updates.append((key, node)) self._registered_updates.add(key)
Register updates that will only be executed in training phase.
def register_training_updates(self, *updates): """ Register updates that will only be executed in training phase. """ for key, node in updates: if key not in self._registered_training_updates: self.training_updates.append((key, node)) self._reg...
Register monitors they should be tuple of name and Theano variable.
def register_monitors(self, *monitors): """ Register monitors they should be tuple of name and Theano variable. """ for key, node in monitors: if key not in self._registered_monitors: node *= 1.0 # Avoid CudaNdarray self.training_monitors.appen...
Get the L2 norm of multiple tensors. This function is taken from blocks.
def multiple_l2_norm(tensors): """ Get the L2 norm of multiple tensors. This function is taken from blocks. """ # Another way for doing this, I don't know which one is fast # return T.sqrt(sum(T.sum(t ** 2) for t in tensors)) flattened = [T.as_tensor_variable(t).flatten() for t in tensors] ...
dumps one element to file_obj a file opened in write mode
def dump_one(elt_to_pickle, file_obj): """ dumps one element to file_obj, a file opened in write mode """ pickled_elt_str = dumps(elt_to_pickle) file_obj.write(pickled_elt_str) # record separator is a blank line # (since pickled_elt_str might contain its own newli...
load contents from file_obj returning a generator that yields one element at a time
def load(file_obj): """ load contents from file_obj, returning a generator that yields one element at a time """ cur_elt = [] for line in file_obj: cur_elt.append(line) if line == '\n': pickled_elt_str = ''.join(cur_elt) cur_el...
Fix the block register all the parameters of sub layers.: return:
def fix(self): """ Fix the block, register all the parameters of sub layers. :return: """ if not self.fixed: for layer in self.layers: if not layer.initialized: raise Exception("All sub layers in a block must be initialized when fix...
Register one connected layer.: type layer: NeuralLayer
def register_layer(self, layer): """ Register one connected layer. :type layer: NeuralLayer """ if self.fixed: raise Exception("After a block is fixed, no more layers can be registered.") self.layers.append(layer)
Load parameters to the block.
def load_params(self, path, exclude_free_params=False): from deepy.core import graph """ Load parameters to the block. """ from deepy.core.comp_graph import ComputationalGraph model = graph.compile(blocks=[self]) model.load_params(path, exclude_free_params=exclude...
Compute one step in the RNN.: return: one variable for RNN and GRU multiple variables for LSTM
def compute_step(self, state, lstm_cell=None, input=None, additional_inputs=None): """ Compute one step in the RNN. :return: one variable for RNN and GRU, multiple variables for LSTM """ if not self.initialized: input_dim = None if input and hasattr(input....
: type input_var: T. var: rtype: dict
def get_initial_states(self, input_var, init_state=None): """ :type input_var: T.var :rtype: dict """ initial_states = {} for state in self.state_names: if state != "state" or not init_state: if self._input_type == 'sequence' and input_var.ndim...
: type input_var: T. var: rtype: dict
def get_step_inputs(self, input_var, states=None, mask=None, additional_inputs=None): """ :type input_var: T.var :rtype: dict """ step_inputs = {} if self._input_type == "sequence": if not additional_inputs: additional_inputs = [] i...