本文整理匯總了Python中Dataset.Dataset方法的典型用法代碼示例。如果您正苦於以下問題：Python Dataset.Dataset方法的具體用法？Python Dataset.Dataset怎麽用？Python Dataset.Dataset使用的例子？那麽, 這裏精選的方法代碼示例或許可以為您提供幫助。您也可以進一步了解該方法所在類Dataset的用法示例。
在下文中一共展示了Dataset.Dataset方法的11個代碼示例，這些例子默認根據受歡迎程度排序。您可以為喜歡或者感覺有用的代碼點讚，您的評價將有助於係統推薦出更棒的Python代碼示例。
示例1: map
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def map(self, mapping_f, filename = None):
if filename is None:
fname = tempfile.mktemp()
delete = True
else:
fname = filename
delete = False
f = h5py.File(fname, "w")
#Transform file by file
for x in self.file_iterator(path = True):
route = x[-1]
x = x[:-1]
y = mapping_f(*x)
if np.prod(y[0].shape) > 0:
for i,v in enumerate(y):
f[route+"/"+str(i)] = v
f.close()
return Dataset(fname,
self.entries_regexp,
tuple(str(i) for i in xrange(len(y))),
delete_after_use = delete) 示例2: _prepare_graph
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def _prepare_graph(self, params):
tf.reset_default_graph()
self.lr = tf.placeholder(tf.float32, shape=())
self.dataset = Dataset(params)
logits = self._prepare_model(self.dataset.img_data)
float_y = tf.cast(self.dataset.labels, tf.float32)
cross_entropy = tf.nn.sigmoid_cross_entropy_with_logits(logits=logits, labels=float_y)
self.loss = tf.reduce_sum(cross_entropy)
sigmoid_logits = tf.nn.sigmoid(logits)
self.predictions = tf.cast(tf.round(sigmoid_logits), tf.int32, name='predictions')
self.accuracy = tf.reduce_sum(tf.reduce_min(tf.cast(tf.equal(self.predictions, self.dataset.labels),
tf.float32), axis=1))
self._prepare_optimizer_stage(fine_tune_upto=1) 示例3: _prepare_graph
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def _prepare_graph(self, params):
tf.reset_default_graph()
self.dataset = Dataset(params)
self.lr = tf.placeholder(tf.float32, shape=())
one_hot_y = tf.one_hot(self.dataset.labels, depth=self._n_class)
logits = self._prepare_model(self.dataset.text_data, self.dataset.text_len)
cross_entropy = tf.nn.softmax_cross_entropy_with_logits_v2(logits=logits, labels=one_hot_y)
self.loss = tf.reduce_sum(cross_entropy)
self.optimizer = tf.train.AdamOptimizer(learning_rate=self.lr).minimize(self.loss)
self.predictions = tf.argmax(logits, axis=1, output_type=tf.int32, name='predictions')
self.accuracy = tf.reduce_sum(tf.cast(tf.equal(self.predictions, self.dataset.labels), tf.float32)) 示例4: _prepare_graph
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def _prepare_graph(self, params):
tf.reset_default_graph()
self.dataset = Dataset(params)
self.lr = tf.placeholder(tf.float32, shape=())
one_hot_y = tf.one_hot(self.dataset.labels, depth=self._n_class)
logits = self._prepare_model(self.dataset.text_data)
cross_entropy = tf.nn.softmax_cross_entropy_with_logits_v2(logits=logits, labels=one_hot_y)
self.loss = tf.reduce_sum(cross_entropy)
self.optimizer = tf.train.AdamOptimizer(learning_rate=self.lr).minimize(self.loss)
self.predictions = tf.argmax(logits, axis=1, output_type=tf.int32, name='predictions')
self.accuracy = tf.reduce_sum(tf.cast(tf.equal(self.predictions, self.dataset.labels), tf.float32)) 示例5: _build_architecture
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def _build_architecture(self, params):
tf.reset_default_graph()
self.dataset = Dataset(params)
self.lr = tf.placeholder(tf.float32, ())
one_hot_y = tf.one_hot(self.dataset.data_y, self._n_class, dtype=tf.int32)
self.logits = self._build_model(self.dataset.data_X)
self.logits = tf.identity(self.logits, name='logits')
self.predictions = tf.argmax(self.logits, axis=1, output_type=tf.int32, name='predictions')
softmax = tf.nn.softmax_cross_entropy_with_logits_v2(labels=one_hot_y, logits=self.logits)
self.loss = tf.reduce_sum(softmax)
self.optimizer = tf.train.RMSPropOptimizer(learning_rate=self.lr).minimize(self.loss)
self.accuracy = tf.reduce_sum(tf.cast(tf.equal(self.predictions, self.dataset.data_y), tf.float32),
name='accuracy') 示例6: _prepare_graph
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def _prepare_graph(self, params):
tf.reset_default_graph()
self.lr = tf.placeholder(tf.float32, shape=())
self.is_training = tf.placeholder(tf.bool, shape=(), name='is_training')
self.dataset = Dataset(params)
logits = self._prepare_model(self.dataset.img_data, self.is_training)
float_y = tf.cast(self.dataset.labels, tf.float32)
cross_entropy = tf.nn.sigmoid_cross_entropy_with_logits(logits=logits, labels=float_y)
self.loss = tf.reduce_sum(cross_entropy)
sigmoid_logits = tf.nn.sigmoid(logits)
self.predictions = tf.cast(tf.round(sigmoid_logits), tf.int32, name='predictions')
self.accuracy = tf.reduce_sum(tf.reduce_min(tf.cast(tf.equal(self.predictions, self.dataset.labels),
tf.float32), axis=1))
self._prepare_optimizer_stage(fine_tune_upto=1) 示例7: reduce
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def reduce(self, reduction_f, initial_accum):
#Reduce file by file
accum = initial_accum
for x in self.file_iterator():
accum = reduction_f(accum, *x)
return accum
# def pmap(self, mapping_f, filename = None, n_workers = 4):
# if filename is None:
# fname = tempfile.mktemp()
# delete = True
# else:
# fname = filename
# delete = False
# f = h5py.File(fname, "w")
# queue = Queue(maxsize = n_workers * 2)
# worker_threads = [MappingThread(queue, mapping_f, f) for i in range(n_workers)]
# # Start all threads
# [t.start() for t in worker_threads]
# # Transform file by file. This is done by queuing it and waiting for the worker threads to do it.
# for x in self.file_iterator(path = True):
# queue.put(x)
# # Send a signal to all threads so they finish
# [t.finish() for t in worker_threads]
# # Wait for all the processing and writing to be finished before closing the file
# queue.join()
# f.close()
# # Return the new dataset
# return Dataset(fname,
# self.entries_regexp,
# tuple(str(i) for i in xrange(len(y))),
# delete_after_use = delete)
#TODO: Move to utils 示例8: _prepare_graph
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def _prepare_graph(self, params):
tf.reset_default_graph()
self.dataset = Dataset(params)
logits = self._prepare_model(self.dataset.img_data)
softmax = tf.nn.softmax(logits)
self.top_prediction = tf.nn.top_k(softmax, self._top_k, name='top_prediction') 示例9: _prepare_graph
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def _prepare_graph(self, params):
tf.reset_default_graph()
self.dataset = Dataset(params)
self.class_entities, self.boxes = self._prepare_model(self.dataset.img_data) 示例10: _build_architecture
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def _build_architecture(self):
tf.reset_default_graph()
self.dataset = Dataset()
self.aug_images = self._build_model() 示例11: map
# 需要導入模塊: import Dataset [as 別名]
# 或者: from Dataset import Dataset [as 別名]
def map(self, mapping_f, filename = None):
if filename is None:
fname = tempfile.mktemp()
delete = True
else:
fname = filename
delete = False
f = h5py.File(fname, "w")
#Transform file by file
for x in self.file_iterator(path = True):
route = x[-1]
x = x[:-1]
y = mapping_f(*x)
if np.prod(y[0].shape) > 0:
for i,v in enumerate(y):
f[route+"/"+str(i)] = v
f.close()
return Dataset(fname,
".*/.*/.*/.*",
tuple(str(i) for i in xrange(len(y))),
delete_after_use = delete)
# def get_data(self, n = None, proportion = None, accept_less = True):
# if n is None:
# total = sum([self.get_file_shape(0, i)[0] for i in xrange(len(self.file_paths))])
# if proportion is not None:
# n = total * proportion
# else:
# n = total
# data = tuple(np.empty((n, self.get_dimensionality(i))) for i in xrange(self.get_arity()))
# row = 0
# for fs in self.file_iterator():
# for i,f in enumerate(fs):
# increment = min(f.shape[0], n-row)
# data[i][row:row+increment, :] = f[0:increment, :]
# row += increment
# if row >= n:
# break
# if accept_less and row < n:
# return tuple(d[0:row,:] for d in data)
# else:
# assert(n == row)
# return data