查看版本
import tensorflow as tf
print tf.__version__
混合模型
| ‘‘‘ | |
| Demonstrate that weights saved with models in one scope, can be loaded | |
| into models being used in a different scope. | |
| This allows multiple models to be run, and combined models to load | |
| weights from separately trained models. | |
| ‘‘‘ | |
| from __future__ import division, print_function, absolute_import | |
| import re | |
| import tflearn | |
| import tensorflow as tf | |
| import tflearn.datasets.mnist as mnist | |
| from tflearn.layers.core import input_data, dropout, fully_connected | |
| from tflearn.layers.conv import conv_2d, max_pool_2d | |
| from tflearn.layers.normalization import local_response_normalization | |
| from tflearn.layers.estimator import regression | |
| #----------------------------------------------------------------------------- | |
| class Model1(object): | |
| ‘‘‘ | |
| convnet MNIST | |
| ‘‘‘ | |
| def __init__(self): | |
| network = tflearn.input_data(shape=[None, 784], name="input") | |
| network = self.make_core_network(network) | |
| network = regression(network, optimizer=‘adam‘, learning_rate=0.01, | |
| loss=‘categorical_crossentropy‘, name=‘target‘) | |
| model = tflearn.DNN(network, tensorboard_verbose=0) | |
| self.model = model | |
| @staticmethod | |
| def make_core_network(network): | |
| network = tflearn.reshape(network, [-1, 28, 28, 1], name="reshape") | |
| network = conv_2d(network, 32, 3, activation=‘relu‘, regularizer="L2") | |
| network = max_pool_2d(network, 2) | |
| network = local_response_normalization(network) | |
| network = conv_2d(network, 64, 3, activation=‘relu‘, regularizer="L2") | |
| network = max_pool_2d(network, 2) | |
| network = local_response_normalization(network) | |
| network = fully_connected(network, 128, activation=‘tanh‘) | |
| network = dropout(network, 0.8) | |
| network = fully_connected(network, 256, activation=‘tanh‘) | |
| network = dropout(network, 0.8) | |
| network = fully_connected(network, 10, activation=‘softmax‘) | |
| return network | |
| def train(self, X, Y, testX, testY, n_epoch=1, snapshot_step=1000): | |
| # Training | |
| self.model.fit({‘input‘: X}, {‘target‘: Y}, n_epoch=n_epoch, | |
| validation_set=({‘input‘: testX}, {‘target‘: testY}), | |
| snapshot_step=snapshot_step, | |
| show_metric=True, run_id=‘convnet_mnist‘) | |
| class Model2(object): | |
| ‘‘‘ | |
| dnn MNIST | |
| ‘‘‘ | |
| def __init__(self): | |
| # Building deep neural network | |
| network = tflearn.input_data(shape=[None, 784], name="input") | |
| network = self.make_core_network(network) | |
| # Regression using SGD with learning rate decay and Top-3 accuracy | |
| sgd = tflearn.SGD(learning_rate=0.1, lr_decay=0.96, decay_step=1000) | |
| top_k = tflearn.metrics.Top_k(3) | |
| network = tflearn.regression(network, optimizer=sgd, metric=top_k, | |
| loss=‘categorical_crossentropy‘, name="target") | |
| model = tflearn.DNN(network, tensorboard_verbose=0) | |
| self.model = model | |
| @staticmethod | |
| def make_core_network(network): | |
| dense1 = tflearn.fully_connected(network, 64, activation=‘tanh‘, | |
| regularizer=‘L2‘, weight_decay=0.001, name="dense1") | |
| dropout1 = tflearn.dropout(dense1, 0.8) | |
| dense2 = tflearn.fully_connected(dropout1, 64, activation=‘tanh‘, | |
| regularizer=‘L2‘, weight_decay=0.001, name="dense2") | |
| dropout2 = tflearn.dropout(dense2, 0.8) | |
| softmax = tflearn.fully_connected(dropout2, 10, activation=‘softmax‘, name="softmax") | |
| return softmax | |
| def train(self, X, Y, testX, testY, n_epoch=1, snapshot_step=1000): | |
| # Training | |
| self.model.fit(X, Y, n_epoch=n_epoch, validation_set=(testX, testY), | |
| snapshot_step=snapshot_step, | |
| show_metric=True, run_id="dense_model") | |
| class Model12(object): | |
| ‘‘‘ | |
| Combination of two networks | |
| ‘‘‘ | |
| def __init__(self): | |
| inputs = tflearn.input_data(shape=[None, 784], name="input") | |
| with tf.variable_scope("scope1") as scope: | |
| net_conv = Model1.make_core_network(inputs) # shape (?, 10) | |
| with tf.variable_scope("scope2") as scope: | |
| net_dnn = Model2.make_core_network(inputs) # shape (?, 10) | |
| network = tf.concat([net_conv, net_dnn], 1, name="concat") # shape (?, 20) | |
| network = tflearn.fully_connected(network, 10, activation="softmax") | |
| network = regression(network, optimizer=‘adam‘, learning_rate=0.01, | |
| loss=‘categorical_crossentropy‘, name=‘target‘) | |
| self.model = tflearn.DNN(network, tensorboard_verbose=0) | |
| def load_from_two(self, m1fn, m2fn): | |
| self.model.load(m1fn, scope_for_restore="scope1", weights_only=True) | |
| self.model.load(m2fn, scope_for_restore="scope2", weights_only=True, create_new_session=False) | |
| def train(self, X, Y, testX, testY, n_epoch=1, snapshot_step=1000): | |
| # Training | |
| self.model.fit(X, Y, n_epoch=n_epoch, validation_set=(testX, testY), | |
| snapshot_step=snapshot_step, | |
| show_metric=True, run_id="model12") | |
| #----------------------------------------------------------------------------- | |
| X, Y, testX, testY = mnist.load_data(one_hot=True) | |
| def prepare_model1_weights_file(): | |
| tf.reset_default_graph() | |
| m1 = Model1() | |
| m1.train(X, Y, testX, testY, 2) | |
| m1.model.save("model1.tfl") | |
| def prepare_model1_weights_file_in_scopeQ(): | |
| tf.reset_default_graph() | |
| with tf.variable_scope("scopeQ") as scope: | |
| m1 = Model1() | |
| m1.model.fit({"scopeQ/input": X}, {"scopeQ/target": Y}, n_epoch=1, validation_set=0.1, show_metric=True, run_id="model1_scopeQ") | |
| m1.model.save("model1_scopeQ.tfl") | |
| def prepare_model2_weights_file(): | |
| tf.reset_default_graph() | |
| m2 = Model2() | |
| m2.train(X, Y, testX, testY, 1) | |
| m2.model.save("model2.tfl") | |
| def demonstrate_loading_weights_into_different_scope(): | |
| print("="*60 + " Demonstrate loading weights saved in scopeQ, into variables now in scopeA") | |
| tf.reset_default_graph() | |
| with tf.variable_scope("scopeA") as scope: | |
| m1a = Model1() | |
| print ("=" * 60 + " Trying to load model1 weights from scopeQ into scopeA") | |
| m1a.model.load("model1_scopeQ.tfl", variable_name_map=("scopeA", "scopeQ"), verbose=True) | |
| def demonstrate_loading_weights_into_different_scope_using_custom_function(): | |
| print("="*60 + " Demonstrate loading weights saved in scopeQ, into variables now in scopeA, using custom map function") | |
| tf.reset_default_graph() | |
| def vname_map(ename): # variables were saved in scopeA, but we want to load into scopeQ | |
| name_in_file = ename.replace("scopeA", "scopeQ") | |
| print ("%s -> %s" % (ename, name_in_file)) | |
| return name_in_file | |
| with tf.variable_scope("scopeA") as scope: | |
| m1a = Model1() | |
| print ("=" * 60 + " Trying to load model1 weights from scopeQ into scopeA") | |
| m1a.model.load("model1_scopeQ.tfl", variable_name_map=vname_map, verbose=True) | |
| def demonstrate_loading_two_instances_of_model1(): | |
| print("="*60 + " Demonstrate loading weights from model1 into two instances of model1 in scopeA and scopeB") | |
| tf.reset_default_graph() | |
| with tf.variable_scope("scopeA") as scope: | |
| m1a = Model1() | |
| print ("-" * 40 + " Trying to load model1 weights: should fail") | |
| try: | |
| m1a.model.load("model1.tfl", weights_only=True) | |
| except Exception as err: | |
| print ("Loading failed, with error as expected, because variables are in scopeA") | |
| print ("error: %s" % str(err)) | |
| print ("-" * 40) | |
| print ("=" * 60 + " Trying to load model1 weights: should succeed") | |
| m1a.model.load("model1.tfl", scope_for_restore="scopeA", verbose=True, weights_only=True) | |
| with tf.variable_scope("scopeB") as scope: | |
| m1b = Model1() | |
| m1b.model.load("model1.tfl", scope_for_restore="scopeB", verbose=True, weights_only=True) | |
| print ("="*60 + " Successfully restored weights to two instances of model1, in different scopes") | |
| def demonstrate_combined_model1_and_model2_network(): | |
| print("="*60 + " Demonstrate loading weights from model1 and model2 into new mashup network model12") | |
| print ("-"*40 + " Creating mashup of model1 and model2 networks") | |
| tf.reset_default_graph() | |
| m12 = Model12() | |
| print ("-"*60 + " Loading model1 and model2 weights into mashup") | |
| m12.load_from_two("model1.tfl", "model2.tfl") | |
| print ("-"*60 + " Training mashup") | |
| m12.train(X, Y, testX, testY, 1) | |
| print ("-"*60 + " Saving mashup weights") | |
| m12.model.save("model12.tfl") | |
| print ("-"*60 + " Done") | |
| print("="*77) | |
| prepare_model1_weights_file() | |
| prepare_model2_weights_file() | |
| prepare_model1_weights_file_in_scopeQ() | |
| print("-"*77) | |
| print("-"*77) | |
| demonstrate_loading_weights_into_different_scope() | |
| demonstrate_loading_weights_into_different_scope_using_custom_function() | |
| demonstrate_loading_two_instances_of_model1() | |
| demonstrate_combined_model1_and_model2_network() | |
| print("="*77) |
时间： 2024-08-29 15:48:52